Nemotron 3 Nano Omni is an open multimodal model built on a hybrid Mixture-of-Experts (MoE) architecture, engineered for high efficiency and strong accuracy across image, video, audio, and text inputs. It powers always-on sub-agents for computer use, document intelligence, and audio-video understanding—replacing fragmented vision, speech, and language pipelines with a single unified inference pass.

Nemotron-3-Nano-Omni-30B-A3B-Reasoning

DeepSeek V4 Pro is an MoE model with 1.6T total parameters (49B active) and a 1M-token context window. It's built for advanced reasoning, coding, and long-running agent tasks, and performs well on knowledge, math, and software engineering benchmarks.

DeepSeek-V4-Pro

DeepSeek V4 Flash is an efficiency-focused MoE model with 284B total parameters (13B active) and a 1M-token context window. It's tuned for fast inference and high-throughput use cases while still holding up on reasoning and coding tasks.

DeepSeek-V4-Flash

Kimi K2.6 is an open-source, native multimodal agentic model that advances practical capabilities in long-horizon coding, coding-driven design, proactive autonomous execution, and swarm-based task orchestration.

Kimi-K2.6

Qwen3.6-35B-A3B is Alibaba's latest flagship Mixture-of-Experts model, with 35B total parameters and only 3B activated per token (256 experts, 8 routed + 1 shared). Built on direct feedback from the community, Qwen3.6 prioritizes stability and real-world utility, offering developers a more intuitive, responsive, and genuinely productive coding experience.

Qwen3.6-35B-A3B

GLM-5.1 is Z-AI's next-generation flagship model for agentic engineering, with significantly stronger coding capabilities than its predecessor. It achieves state-of-the-art performance on SWE-Bench Pro and leads GLM-5 by a wide margin on NL2Repo (repo generation) and Terminal-Bench 2.0 (real-world terminal tasks).

GLM-5.1

Step 3.5 Flash is an open-source reasoning model by StepFun with 196B total parameters (11B active) using Mixture of Experts. It features a 256K context window, deep reasoning, tool calling, and agentic capabilities, achieving 97.3 on AIME 2025 and 74.4% on SWE-bench Verified.

Step-3.5-Flash

Qwen3.5-397B-A17B is Alibaba's most capable Qwen3.5 model, a Mixture-of-Experts architecture with 397B total parameters and 17B activated per token. It features a 262K token context window (extensible to 1M with YaRN), thinking/reasoning mode, tool calling with MCP integration, and support for 201 languages. Sets state-of-the-art results on reasoning, coding, math, and multimodal benchmarks.

Qwen3.5-397B-A17B

Efficient, MoE variant of Gemma 4. Gemma is a family of open models built by Google DeepMind. Gemma 4 models are multimodal, handling text and image input and generating text output.

gemma-4-26B-A4B-it

Gemma is a family of open models built by Google DeepMind. Gemma 4 models are multimodal, handling text and image input and generating text output.

gemma-4-31B-it

Qwen3.5-122B-A10B is a large Mixture-of-Experts model from Alibaba's Qwen3.5 series with 122B total parameters and 10B activated per token. It features a 262K token context window (extensible to 1M with YaRN), thinking/reasoning mode, tool calling, and support for 201 languages. Excels at complex reasoning, coding, multimodal understanding, and agentic tasks with the efficiency of sparse activation.

Qwen3.5-122B-A10B

NVIDIA Nemotron 3 Super is a hybrid Mixture-of-Experts (MoE) model engineered for highest compute efficiency and accuracy in multi-agent applications and specialized agentic systems. It is optimized to run many collaborating agents per application on a single GPU, delivering high accuracy for reasoning, tool use, and instruction following.

NVIDIA-Nemotron-3-Super-120B-A12B

GLM-5 is an advanced, open-source large language model designed for developers tackling the toughest challenges. It excels at long-context reasoning, multi-step tool orchestration, and complex systems engineering, making it the ideal choice for powering sophisticated agents and applications that require high-level cognitive tasks.

GLM-5

  Qwen3-TTS is an advanced text-to-speech model by Alibaba's Qwen team, delivering stable, expressive, and low-latency speech generation across 10 languages.                                                                                                                                                                                                                                                                                                                                           Key capabilities:                                                                                                                                                                                                                                  - 9 preset voices — Vivian, Serena, Uncle_Fu, Dylan, Eric, Ryan, Aiden, Ono_Anna, Sohee — covering diverse genders, ages, and accents                                                                                                              - Voice cloning — clone any voice from a short (~3s) audio sample via the voice_id parameter   - Instruction control — adjust tone, emotion, and speaking style with natural language (e.g. "speak slowly and calmly", "excited tone")   - 10 languages — English, Chinese, Japanese, Korean, German, French, Russian, Spanish, Italian, Portuguese   - Streaming support — real-time PCM streaming with ~97ms first-byte latency   - Multiple output formats — WAV, MP3, FLAC, PCM    Built on a 1.7B parameter architecture using discrete multi-codebook language modeling for end-to-end speech synthesis without cascading errors. Uses a custom 12Hz acoustic tokenizer that preserves paralinguistic information and   environmental audio details.

Qwen3-TTS

● Qwen3-TTS-VoiceDesign is a voice design variant of Qwen3-TTS by Alibaba's Qwen team. Instead of selecting from preset voices, you describe the voice you want in natural language — and the model generates speech in that voice.                                                                                                                                                                                                                                                                     Key capabilities:                                                                                                                                                                                                                                  - Natural language voice control — describe any voice with free text (e.g. "a deep male voice with a calm, authoritative presence", "a young cheerful female with a warm and friendly tone")   - 10 languages — English, Chinese, Japanese, Korean, German, French, Russian, Spanish, Italian, Portuguese                                                                                                                                         - Streaming support — real-time PCM streaming   - Multiple output formats — WAV, MP3, FLAC, PCM    Built on the same 1.7B parameter architecture as Qwen3-TTS, using discrete multi-codebook language modeling and a custom 12Hz acoustic tokenizer for high-quality end-to-end speech synthesis.

Qwen3-TTS-VoiceDesign

MiniMax M2.5 is SOTA in coding, agentic tool use and search, office work, and a range of other economically valuable tasks, boasting scores of 80.2% in SWE-Bench Verified, 51.3% in Multi-SWE-Bench, and 76.3% in BrowseComp (with context management).

MiniMax-M2.5

The latest flagship model in the Qwen family. State-of-the-art results across a comprehensive suite of benchmarks — including knowledge, reasoning, coding, instruction following, human preference alignment, agent tasks, and multilingual understanding.

Qwen3-Max

The latest flagship reasoning model in the Qwen3 family. Further enhanced by multiple innovations like adaptive tool-use and advanced test-time scaling techniques

Qwen3-Max-Thinking

Kimi K2.5 is an open-source, native multimodal agentic model built through continual pretraining on approximately 15 trillion mixed visual and text tokens atop Kimi-K2-Base. It seamlessly integrates vision and language understanding with advanced agentic capabilities, instant and thinking modes, as well as conversational and agentic paradigms.

Kimi-K2.5

GLM-4.7-Flash is a 30B-A3B MoE model. As the strongest model in the 30B class, GLM-4.7-Flash offers a new option for lightweight deployment that balances performance and efficiency.

GLM-4.7-Flash

DeepSeek-V3.2 is a large language model designed to harmonize high computational efficiency with strong reasoning and agentic tool-use performance. It introduces DeepSeek Sparse Attention (DSA), a fine-grained sparse attention mechanism that reduces training and inference cost while preserving quality in long-context scenarios. A scalable reinforcement learning post-training framework further improves reasoning, with reported performance in the GPT-5 class, and the model has demonstrated gold-medal results on the 2025 IMO and IOI. V3.2 also uses a large-scale agentic task synthesis pipeline to better integrate reasoning into tool-use settings, boosting compliance and generalization in interactive environments.

DeepSeek-V3.2

The fastest model of the Flux 2 family. Frontier visual intelligence — state-of-the-art image generation and editing from Black Forest Labs

FLUX-2-klein-4b

The best quality-to-latency ratio, production apps model of the Flux 2 family. Frontier visual intelligence — state-of-the-art image generation and editing from Black Forest Labs

FLUX-2-klein-9b

Developed by Anthropic, Claude is a family of highly performant, trustworthy AI models built for complex reasoning, advanced coding, and nuanced language understanding. The latest Claude 4 generation delivers breakthrough capabilities in analytical thinking, with Claude 4 Opus setting new standards for intelligence and Claude 4 Sonnet providing exceptional performance with remarkable efficiency.

Claude models excel at understanding context, following complex instructions, and maintaining coherent conversations across extended interactions. With advanced features like extended thinking for deeper reasoning, prompt caching that reduces costs by up to 90%, vision capabilities for image analysis, and robust safety measures, Claude is designed for enterprise applications that demand both sophistication and reliability.

Available with comprehensive API features including streaming responses, batch processing for 50% cost savings, multilingual support across dozens of languages, and flexible context windows up to 200K tokens (1M in beta), Claude is perfect for building intelligent applications like customer support agents, content analysis systems, coding assistants, and complex reasoning workflows that require both accuracy and trustworthiness.

Claude AI family: Claude 4 Opus for complex reasoning, Claude 4 Sonnet for balanced performance, plus advanced capabilities like extended thinking, prompt caching, vision analysis, and enterprise-grade safety APIs.

Claude AI APIs via DeepInfra

claude

Developed by Anthropic, Claude is a family of highly performant, trustworthy AI models built for complex reasoning, advanced coding, and nuanced language understanding

DeepInfra provides access to Anthropic's latest Claude models, featuring the most advanced reasoning capabilities and balanced performance options, all with enterprise-grade safety and reliability.

Claude

DeepSeek develops advanced foundation models optimized for computational efficiency and strong generalization across diverse tasks. The architecture incorporates recent advances in transformer-based systems, delivering robust performance in both zero-shot and fine-tuned scenarios. Models are pretrained on rigorously filtered multilingual corpora with specialized optimizations for mathematical reasoning and algorithmic tasks. The inference stack achieves competitive throughput while maintaining low latency, making it suitable for production deployment. Researchers and engineers can leverage these models for tasks ranging from natural language processing to complex analytical problem-solving.

deepseek

DeepSeek's models are a suite of advanced AI systems that prioritize efficiency, scalability, and real-world applicability.

DeepSeek

Developed by Black Forest Labs (the original creators behind Stable Diffusion), Flux is a family of state-of-the-art image generation and editing models that deliver exceptional visual quality with breakthrough prompt accuracy and photorealism. Built on advanced 12 billion parameter architecture, Flux models excel at understanding exactly what you want to create or modify.
FLUX 2 represents the next generation of the Flux family, introducing significantly improved image quality, faster generation, and more precise prompt following. The lineup includes Max for ultimate quality, Pro for production workflows, and Klein variants (9B and 4B) optimized for speed and efficiency — making high-quality image generation accessible at every scale.
Flux offers specialized variants for every need: from open-source to commercial licensing, Flux is perfect for developers building creative applications, product visualization tools, and next-generation image editing experiences.

Flux AI image generation family: FLUX.1 Kontext for in-context editing, FLUX.1 Pro/Dev for text-to-image synthesis, plus comprehensive editing tools and state-of-the-art visual generation APIs.

Flux Image Generation APIs via DeepInfra

flux

Developed by Black Forest Labs, Flux is a family of state-of-the-art image generation and editing models that deliver exceptional visual quality with breakthrough prompt accuracy and photorealism.

DeepInfra provides access to Black Forest Labs' complete Flux ecosystem, offering everything from lightning-fast generation to sophisticated in-context editing capabilities with industry-leading prompt adherence and visual quality.

Flux

Developed by Google DeepMind, Gemini is a family of state-of-the-art thinking models with native multimodal capabilities, designed for advanced reasoning, complex problem-solving, and comprehensive understanding across text, audio, video, and images. Built with revolutionary thinking architecture, Gemini models reason through problems step-by-step before responding, delivering enhanced accuracy and performance for sophisticated applications.

Gemini 2.5 Pro sets new standards for complex reasoning and coding excellence, while Gemini 2.5 Flash provides optimal price-performance for high-volume tasks. With massive context windows up to 1 million tokens, native multimodal processing that handles hours of video and audio, and transparent reasoning capabilities that show step-by-step thinking processes, Gemini excels at document analysis, code generation, scientific research, and agentic workflows.

Perfect for building intelligent applications that require deep reasoning, multimodal understanding, long-context processing, and transparent AI decision-making with Google's enterprise-grade reliability and performance.


Gemini AI family: Advanced thinking models with native multimodal processing for text, audio, video, and image understanding APIs

Gemini AI Model APIs via DeepInfra

gemini

Developed by Google DeepMind, Gemini is a family of state-of-the-art thinking models with native multimodal capabilities

DeepInfra provides access to Google's latest Gemini models, featuring advanced thinking capabilities, native multimodal processing, and industry-leading performance for complex reasoning and development tasks.

Gemini

Developed by Meta, Llama (Large Language Model Meta AI) is a family of state-of-the-art open-weight models designed for efficiency and performance. The latest versions feature Mixture-of-Experts (MoE) architectures, enabling cost-effective inference by dynamically activating subsets of parameters. With support for multimodal inputs (text + images) and extended context windows (up to 10M tokens), Llama excels in tasks like code generation, multilingual understanding, and long-form reasoning. The models support FP8 quantization and batch inference, ensuring low-latency, high-throughput performance for production workloads. With permissive licensing and robust tooling (e.g., Llama Guard for safety), Llama is ideal for developers seeking powerful, customizable AI with minimal overhead.

llama

The Llama 4 collection of models are natively multimodal AI models that enable text and multimodal experiences.

Llama 4

Meta Llama 3 are a collection of pretrained and instruction tuned generative text models in 8B, 70B and 405B sizes.

Llama 3

Llama

Developed by Mistral AI, a leading French research lab, Mistral is a family of open-source AI models built for multilingual excellence, advanced reasoning, and cost-effective performance. These models excel at complex reasoning, mathematics, coding, and specialized tasks while offering complete transparency and deployment freedom through open-source licensing.

Mistral Small 3.2 delivers breakthrough efficiency with native fluency in European languages, while specialized variants handle specific needs: Devstral for coding, Voxtral for audio processing, and Mixtral for high-performance tasks. With Apache 2.0 licensing, extensive context windows up to 128K tokens, and comprehensive customization options, Mistral provides enterprise-grade capabilities without vendor lock-in.

Perfect for building multilingual applications, coding assistants, and reasoning systems where you need both powerful performance and complete control over your AI deployment.

Mistral AI model family: Mistral Small 3.2 for efficient performance, Mixtral for specialized tasks, Devstral for coding, plus multilingual reasoning, mathematics, and open-source flexibility APIs.

Mistral AI Model APIs via DeepInfra

mistral

Developed by Mistral AI, a leading French research lab, Mistral is a family of open-source AI models built for multilingual excellence, advanced reasoning, and cost-effective performance

DeepInfra provides access to Mistral AI's comprehensive open-source model ecosystem, from efficient small models to specialized coding and audio processing variants, all with complete Apache 2.0 licensing freedom.

Mistral

Voxtral is a family of audio models with state-of-the-art speech to text capabilities.

Voxtral

The Nemotron family spans Omni, Nano, Super, and specialized instruct variants, enabling you to
  balance accuracy, reasoning depth, latency, and cost for your specific workload.
  Omni for multimodal reasoning across text, audio, and video
  Nano for maximum efficiency and stable inference
  Super for multi-agent systems and advanced reasoning
  Instruct variants for instruction-following and conversational workloads

nemotron

[NVIDIA Nemotron™](https://www.nvidia.com/en-us/ai-data-science/foundation-models/nemotron/) is a
  family of open models, datasets, and training recipes engineered for high performance, efficiency, and customization.
   Nemotron models support synthetic data workflows and supervised fine-tuning—and are optimized for real-time
  inference, reasoning agents, and production AI systems.

The Nemotron family spans Nano, Super, and specialized instruct variants, enabling you to balance accuracy, reasoning depth, latency, and cost for your specific workload.
- **Nano** for maximum efficiency and stable inference
- **Super** for multi-agent systems and advance reasoning
- **Instruct variants** for instruction-following and conversational workloads

Nemotron

Developed by Alibaba Group's Qwen Team, Qwen is a family of state-of-the-art large language and multimodal models designed for comprehensive AI capabilities and multilingual performance. The latest Qwen3 generation features balanced model architectures including reintroduced Mixture-of-Experts (MoE) variants (Qwen3-30B-A3B and Qwen3-235B-A22B) alongside dense models up to 32B parameters, enabling efficient resource utilization through dynamic parameter activation. 

With support for 119 languages and dialects, hybrid thinking modes that seamlessly alternate between reasoning and instruction-following without model switching, and extended context windows (up to 1M tokens in Qwen3-2507), Qwen excels in tasks like multilingual understanding, code generation, agentic workflows, and complex problem-solving. The models utilize advanced Byte-level Byte Pair Encoding with a 151,646-token vocabulary, structured ChatML formatting for conversational interactions, and robust tool calling capabilities with parallel execution support. 

Available in both proprietary and open-weight versions with flexible licensing, comprehensive model variants (Base, Instruct, Thinking, and hybrid modes), and enhanced Model Context Protocol support, Qwen is ideal for developers seeking powerful, multilingual AI systems with sophisticated reasoning capabilities and minimal deployment complexity.

Qwen AI model family: Qwen3 language models, specialized coding & reasoning models, plus state-of-the-art embedding & reranking APIs for search and RAG applications.

Qwen Model APIs via DeepInfra

qwen

Qwen series offers a comprehensive suite of dense and mixture-of-experts models.

DeepInfra provides access to Qwen's latest generation of large language models, offering both specialized coding models and general-purpose AI systems with advanced reasoning capabilities.

Qwen

Meet Qwen3-VL — the most powerful vision-language model in the Qwen series to date.  This generation delivers comprehensive upgrades across the board: superior text understanding & generation, deeper visual perception & reasoning, extended context length, enhanced spatial and video dynamics comprehension, and stronger agent interaction capabilities.

MiniMax-M2 is a Mini model built for Max coding & agentic workflows with just 10 billion activated parameters

Stable Diffusion is a latent text-to-image diffusion model. Generate realistic images given text description

GLM-4.7 is a state-of-the-art, multilingual Mixture-of-Experts (MoE) language model designed for complex reasoning, agentic coding, and tool use. Building on its predecessor GLM-4.6, it delivers significant improvements across key benchmarks, including multilingual SWE-bench, Terminal Bench, and reasoning-heavy evaluations like HLE. The model features advanced "Interleaved Thinking" and new "Preserved Thinking" modes, allowing it to reason before actions and maintain consistency across long, multi-turn tasks. With 358 billion parameters, GLM-4.7 excels in generating clean code, modern UI elements, and sophisticated reasoning outputs.

BGE-M3 is a versatile text embedding model that supports multi-functionality, multi-linguality, and multi-granularity, allowing it to perform dense retrieval, multi-vector retrieval, and sparse retrieval in over 100 languages and with input sizes up to 8192 tokens. The model can be used in a retrieval pipeline with hybrid retrieval and re-ranking to achieve higher accuracy and stronger generalization capabilities. BGE-M3 has shown state-of-the-art performance on several benchmarks, including MKQA, MLDR, and NarritiveQA, and can be used as a drop-in replacement for other embedding models like DPR and BGE-v1.5.

Dolphin 2.9.1, a fine-tuned Llama-3-70b model. The new model, trained on filtered data, is more compliant but uncensored. It demonstrates improvements in instruction, conversation, coding, and function calling abilities.

Kokoro is an open-weight TTS model with 82 million parameters. Despite its lightweight architecture, it delivers comparable quality to larger models while being significantly faster and more cost-efficient. With Apache-licensed weights, Kokoro can be deployed anywhere from production environments to personal projects.

Qwen3.5-4B is a mid-size model from Alibaba's Qwen3.5 series that delivers a strong balance of performance and efficiency. It features a 262K token context window (extensible to 1M with YaRN), thinking/reasoning mode, tool calling, and support for 201 languages. Well-suited for complex reasoning, code generation, and agentic applications.

The Wan2.1 1.3B model is a lightweight, efficient text-to-video generator. Despite its compact size, it delivers impressive performance across benchmarks and generates high-quality 480P videos.

We introduce StarCoder2-15B-Instruct-v0.1, the very first entirely self-aligned code Large Language Model (LLM) trained with a fully permissive and transparent pipeline. Our open-source pipeline uses StarCoder2-15B to generate thousands of instruction-response pairs, which are then used to fine-tune StarCoder-15B itself without any human annotations or distilled data from huge and proprietary LLMs.

NVIDIA Nemotron 2 Nano VL extends the Nemotron family into multi-modal reasoning and document intelligence. This auto-regressive vision-language model enables multi-image reasoning, video understanding, visual Q&A and document analysis and summarization. Optimized for enterprise AI workflows, it powers multimodal agentic systems such as visual copilots, document assistants, and knowledge automation pipelines.

The Qwen3 Embedding model series is the latest proprietary model of the Qwen family, specifically designed for text embedding and ranking tasks. Building upon the dense foundational models of the Qwen3 series, it provides a comprehensive range of text embeddings and reranking models in various sizes (0.6B, 4B, and 8B)

Meta developed and released the Meta Llama 3.1 family of large language models (LLMs), a collection of pretrained and instruction tuned generative text models in 8B, 70B and 405B sizes

High-quality multilingual text-to-speech model by Inworld AI with 130+ preset voices across 15 languages. Supports voice cloning, word-level timestamps, and streaming. Optimized for natural, expressive speech with <250ms time-to-first-audio.

DeepSeek-Prover-V2, an open-source large language model designed for formal theorem proving in Lean 4, with initialization data collected through a recursive theorem proving pipeline powered by DeepSeek-V3. The cold-start training procedure begins by prompting DeepSeek-V3 to decompose complex problems into a series of subgoals. The proofs of resolved subgoals are synthesized into a chain-of-thought process, combined with DeepSeek-V3's step-by-step reasoning, to create an initial cold start for reinforcement learning. 

Kimi K2 Thinking is the latest, most capable version of open-source thinking model developed by MoonshotAI

Hermes 3 is a generalist language model with many improvements over Hermes 2, including advanced agentic capabilities, much better roleplaying, reasoning, multi-turn conversation, long context coherence, and improvements across the board.

Qwen2.5-Coder-7B is a powerful code-specific large language model with 7.61 billion parameters. It's designed for code generation, reasoning, and fixing tasks. The model covers 92 programming languages and has been trained on 5.5 trillion tokens of data, including source code, text-code grounding, and synthetic data.

A sentence transformation model that has been trained on a wide range of datasets, including but not limited to S2ORC, WikiAnwers, PAQ, Stack Exchange, and Yahoo! Answers. Our model can be used for various NLP tasks such as clustering, sentiment analysis, and question answering.

BGE embedding is a general Embedding Model. It is pre-trained using retromae and trained on large-scale pair data using contrastive learning. Note that the goal of pre-training is to reconstruct the text, and the pre-trained model cannot be used for similarity calculation directly, it needs to be fine-tuned

olmOCR is a specialized AI tool that converts PDF documents into clean, structured text while preserving important formatting and layout information. What makes olmOCR particularly valuable for developers is its ability to handle challenging PDFs that traditional OCR tools struggle with—including complex layouts, poor-quality scans, handwritten text, and documents with mixed content types. Built on a fine-tuned 7B vision-language model, olmOCR provides enterprise-grade PDF processing at a fraction of the cost of proprietary solutions.

DeepSeek-V3-0324, a strong Mixture-of-Experts (MoE) language model with 671B total parameters with 37B activated for each token, an improved iteration over DeepSeek-V3.

Phi-4 is a model built upon a blend of synthetic datasets, data from filtered public domain websites, and acquired academic books and Q&A datasets. The goal of this approach was to ensure that small capable models were trained with data focused on high quality and advanced reasoning.

Qwen3-Coder-30B-A3B-Instruct is a high-performance code generation model optimized for agentic coding and complex programming tasks. With 30.5B total parameters and 3.3B activated through Mixture-of-Experts architecture, it delivers exceptional efficiency. The model features native support for 256K token context (extendable to 1M), making it ideal for repository-scale code understanding. It excels at tool calling, browser automation, and multi-step coding workflows.

The Qwen3 Embedding model series is the latest proprietary model of the Qwen family, specifically designed for text embedding and ranking tasks. Building upon the dense foundational models of the Qwen3 series, it provides a comprehensive range of text embeddings and reranking models in various sizes (0.6B, 4B, and 8B).

The Meta Llama 3.2 collection of multilingual large language models (LLMs) is a collection of pretrained and instruction-tuned generative models in 1B and 3B sizes (text in/text out)

Llama-3.1-Nemotron-70B-Instruct is a large language model customized by NVIDIA to improve the helpfulness of LLM generated responses to user queries. This model reaches Arena Hard of 85.0, AlpacaEval 2 LC of 57.6 and GPT-4-Turbo MT-Bench of 8.98, which are known to be predictive of LMSys Chatbot Arena Elo.  As of 16th Oct 2024, this model is #1 on all three automatic alignment benchmarks (verified tab for AlpacaEval 2 LC), edging out strong frontier models such as GPT-4o and Claude 3.5 Sonnet.

The DeepSeek R1 model has undergone a minor version upgrade, with the current version being DeepSeek-R1-0528.

ByteDance's Seedance 1.5 Pro is a professional video model using V2A native generation for integrated, synced audio-visual output, enhancing efficiency of professional video creation.

Compared with GLM-4.5, GLM-4.6 brings several key improvements:  Longer context window: The context window has been expanded from 128K to 200K tokens, enabling the model to handle more complex agentic tasks. Superior coding performance: The model achieves higher scores on code benchmarks and demonstrates better real-world performance in applications such as Claude Code、Cline、Roo Code and Kilo Code, including improvements in generating visually polished front-end pages. Advanced reasoning: GLM-4.6 shows a clear improvement in reasoning performance and supports tool use during inference, leading to stronger overall capability. More capable agents: GLM-4.6 exhibits stronger performance in tool using and search-based agents, and integrates more effectively within agent frameworks. Refined writing: Better aligns with human preferences in style and readability, and performs more naturally in role-playing scenarios.

This is the instruction fine-tuned version of Mixtral-8x22B - the latest and largest mixture of experts large language model (LLM) from Mistral AI. This state of the art machine learning model uses a mixture 8 of experts (MoE) 22b models. During inference 2 experts are selected. This architecture allows large models to be fast and cheap at inference.

The GLM-4.5 series models are foundation models designed for intelligent agents. GLM-4.5 has 355 billion total parameters with 32 billion active parameters, while GLM-4.5-Air adopts a more compact design with 106 billion total parameters and 12 billion active parameters. GLM-4.5 models unify reasoning, coding, and intelligent agent capabilities to meet the complex demands of intelligent agent applications.

DeepSeek-V3.2-Exp is an intermediate step toward the next-generation architecture of the DeepSeek models by introducing DeepSeek Sparse Attention—a sparse attention mechanism designed to explore and validate optimizations for training and inference efficiency in long-context scenarios.

PaddleOCR-VL is a SOTA and resource-efficient model tailored for document parsing. Its core component is PaddleOCR-VL-0.9B, a compact yet powerful vision-language model (VLM) that integrates a NaViT-style dynamic resolution visual encoder with the ERNIE-4.5-0.3B language model to enable accurate element recognition. This innovative model efficiently supports 109 languages and excels in recognizing complex elements (e.g., text, tables, formulas, and charts), while maintaining minimal resource consumption. Through comprehensive evaluations on widely used public benchmarks and in-house benchmarks, PaddleOCR-VL achieves SOTA performance in both page-level document parsing and element-level recognition. It significantly outperforms existing solutions, exhibits strong competitiveness against top-tier VLMs, and delivers fast inference speeds. These strengths make it highly suitable for practical deployment in real-world scenarios.

Qwen3 is the latest generation of large language models in Qwen series, offering a comprehensive suite of dense and mixture-of-experts (MoE) models. Built upon extensive training, Qwen3 delivers groundbreaking advancements in reasoning, instruction-following, agent capabilities, and multilingual support

L3.3-70B-Euryale-v2.3 is a model focused on creative roleplay from Sao10k

Wan2.6 text to image, Upgraded visual quality, aesthetics, and instruction-following deliver precise style control, realistic portraits, long-text understanding, and broad historical/cultural IP coverage, enabling high-quality, highly expressive visual generation.

The GTE models are trained by Alibaba DAMO Academy. They are mainly based on the BERT framework and currently offer three different sizes of models, including GTE-large, GTE-base, and GTE-small. The GTE models are trained on a large-scale corpus of relevance text pairs, covering a wide range of domains and scenarios. This enables the GTE models to be applied to various downstream tasks of text embeddings, including information retrieval, semantic textual similarity, text reranking, etc.

HiggsAudioV2.5 is a high-quality neural text-to-speech (TTS) model designed for natural-sounding voice generation across a wide range of use cases. It focuses on clarity, stable prosody, and consistent pacing, making it suitable for both short prompts and longer narration.

A sentence similarity model that can be used for various NLP tasks such as text classification, sentiment analysis, named entity recognition, question answering, and more. It utilizes the CoSENT architecture, which consists of a transformer encoder and a pooling module, to encode input texts into vectors that capture their semantic meaning. The model was trained on the nli_zh dataset and achieved high performance on various benchmark datasets.

gpt-oss-20b is an open-weight 21B parameter model released by OpenAI under the Apache 2.0 license. It uses a Mixture-of-Experts (MoE) architecture with 3.6B active parameters per forward pass, optimized for lower-latency inference. The model is trained in OpenAI’s Harmony response format and supports reasoning level configuration, fine-tuning, and agentic capabilities including function calling, tool use, and structured outputs.

This model is part of the GLM-V family of models, introduced in the paper GLM-4.1V-Thinking and GLM-4.5V: Towards Versatile Multimodal Reasoning with Scalable Reinforcement Learning.

Whisper is a set of multi-lingual, robust speech recognition models trained by OpenAI that achieve state-of-the-art results in many languages. Whisper models were trained to predict approximate timestamps on speech segments (most of the time with 1-second accuracy), but they cannot originally predict word timestamps. This variant contains implementation to predict word timestamps and provide a more accurate estimation of speech segments when transcribing with Whisper models.

Turn any image into a video. Intelligent shot scheduling supports multi-shot storytelling, generating multi-shot narrative videos with consistent subjects, scenes, and atmosphere

Qwen3-Coder-480B-A35B-Instruct is the Qwen3's most agentic code model, featuring Significant Performance on Agentic Coding, Agentic Browser-Use and other foundational coding tasks, achieving results comparable to Claude Sonnet.

Qwen2.5-Coder is the latest series of Code-Specific Qwen large language models (formerly known as CodeQwen). It has significant improvements in code generation, code reasoning and code fixing. A more comprehensive foundation for real-world applications such as Code Agents. Not only enhancing coding capabilities but also maintaining its strengths in mathematics and general competencies.

Bria 3.2 is the next-generation commercial-ready text-to-image model. With just 4 billion parameters, it provides exceptional aesthetics and text rendering, evaluated to be on par to leading open-source models, and outperforming other licensed models.

Phi-4-reasoning-plus is a state-of-the-art open-weight reasoning model finetuned from Phi-4 using supervised fine-tuning on a dataset of chain-of-thought traces and reinforcement learning. The supervised fine-tuning dataset includes a blend of synthetic prompts and high-quality filtered data from public domain websites, focused on math, science, and coding skills as well as alignment data for safety and Responsible AI. The goal of this approach was to ensure that small capable models were trained with data focused on high quality and advanced reasoning. Phi-4-reasoning-plus has been trained additionally with Reinforcement Learning, hence, it has higher accuracy but generates on average 50% more tokens, thus having higher latency.

Bria Blur Background softens and de-emphasizes image backgrounds while keeping the subject sharp and clear for professional-quality results. Trained fully on licensed data, it delivers safe, natural, and commercial-ready outputs.

Mistral Small 3.1 (2503) adds state-of-the-art vision understanding and extends context capabilities up to 128K tokens while maintaining top-tier text performance. Its 24 billion parameters and instruction fine-tuning deliver fast, local deployment for both text and vision tasks.

Whisper is a pre-trained model for automatic speech recognition (ASR) and speech translation. It was trained on 680k hours of labelled data and demonstrated a strong ability to generalise to many datasets and domains without fine-tuning. Whisper checks pens are available in five configurations of varying model sizes, including a smallest configuration trained on English-only data and a largest configuration trained on multilingual data. This one is English-only.

Llama Guard 3 is a Llama-3.1-8B pretrained model, fine-tuned for content safety classification. Similar to previous versions, it can be used to classify content in both LLM inputs (prompt classification) and in LLM responses (response classification). It acts as an LLM – it generates text in its output that indicates whether a given prompt or response is safe or unsafe, and if unsafe, it also lists the content categories violated.

Euryale 70B v2.1 is a model focused on creative roleplay from Sao10k

Remove unwanted objects or regions from video using a mask, reconstructs the background with intelligent content-aware fill.

Qwen3-235B-A22B-Thinking-2507 is the Qwen3's new model with scaling the thinking capability of Qwen3-235B-A22B, improving both the quality and depth of reasoning. 

Whisper is a pre-trained model for automatic speech recognition (ASR) and speech translation. Trained on 680k hours of labelled data, Whisper models demonstrate a strong ability to generalise to many datasets and domains without fine-tuning. The primary intended users of these models are AI researchers studying robustness, generalisation, and capabilities of the current model.

WizardLM-2 7B is the smaller variant of Microsoft AI's latest Wizard model. It is the fastest and achieves comparable performance with existing 10x larger open-source leading models

Orpheus TTS is a state-of-the-art, Llama-based Speech-LLM designed for high-quality, empathetic text-to-speech generation. This model has been finetuned to deliver human-level speech synthesis, achieving exceptional clarity, expressiveness, and real-time streaming performances.

Gemini 2.5 Flash is Google's latest thinking model, designed to tackle increasingly complex problems. It's capable of reasoning through their thoughts before responding, resulting in enhanced performance and improved accuracy.  Gemini 2.5 Flash: best for balancing reasoning and speed.

Compared with the Plus series, it significantly reduces the “AI-like” feel in generated images, enhancing their realism. It delivers more lifelike material textures for human subjects, finer and more detailed natural textures, and more visually appealing text rendering.

FLUX.1 Redux [dev] is an image variation generation adapter for all FLUX.1 base models. It enables users to refine images with slight variations and supports text-based restyling via API. Integrated with FLUX1.1 [pro] Ultra, it allows for high-quality 4-megapixel outputs. The model can be used with Diffusers in Python for efficient image generation. While powerful, it has ethical and factual limitations and is governed by a non-commercial license.

Black Forest Labs' latest state-of-the art proprietary model sporting top of the line prompt following, visual quality, details and output diversity.

Light and fast. Remove the background of your videos to bring the foreground elements to focus. No more unwanted distractions.

Anthropic’s most powerful model yet and the state-of-the-art coding model. It delivers sustained performance on long-running tasks that require focused effort and thousands of steps, significantly expanding what AI agents can solve. Claude Opus 4 is ideal for powering frontier agent products and features.

DeepSeek-V3.1 is post-trained on the top of DeepSeek-V3.1-Base, which is built upon the original V3 base checkpoint through a two-phase long context extension approach, following the methodology outlined in the original DeepSeek-V3 report. We have expanded our dataset by collecting additional long documents and substantially extending both training phases. The 32K extension phase has been increased 10-fold to 630B tokens, while the 128K extension phase has been extended by 3.3x to 209B tokens. Additionally, DeepSeek-V3.1 is trained using the UE8M0 FP8 scale data format to ensure compatibility with microscaling data formats.

The Llama 4 collection of models are natively multimodal AI models that enable text and multimodal experiences. These models leverage a mixture-of-experts architecture to offer industry-leading performance in text and image understanding. Llama 4 Scout, a 17 billion parameter model with 16 experts

Voxtral Mini is an enhancement of Ministral 3B, incorporating state-of-the-art audio input capabilities while retaining best-in-class text performance. It excels at speech transcription, translation and audio understanding.

  At 2.5 billion parameters, with improved MMDiT-X architecture and training methods, this model is designed to run “out of the box” on consumer hardware, striking a balance between quality and ease of customization. It is capable of generating images ranging between 0.25 and 2 megapixel resolution. 

Accurately preserve the look and voice of people or objects from a reference video, supporting multi-reference co-creation.

DeepSeek R1 Distill Qwen 32B is a distilled large language model based on Qwen 2.5 32B, using outputs from DeepSeek R1. It outperforms OpenAI's o1-mini across various benchmarks, achieving new state-of-the-art results for dense models.  Other benchmark results include:  AIME 2024: 72.6 | MATH-500: 94.3 | CodeForces Rating: 1691.

The Llama 90B Vision model is a top-tier, 90-billion-parameter multimodal model designed for the most challenging visual reasoning and language tasks. It offers unparalleled accuracy in image captioning, visual question answering, and advanced image-text comprehension. Pre-trained on vast multimodal datasets and fine-tuned with human feedback, the Llama 90B Vision is engineered to handle the most demanding image-based AI tasks.  This model is perfect for industries requiring cutting-edge multimodal AI capabilities, particularly those dealing with complex, real-time visual and textual analysis.

We introduce DeepSeek-R1, which incorporates cold-start data before RL. DeepSeek-R1 achieves performance comparable to OpenAI-o1 across math, code, and reasoning tasks. 

MiMo-V2.5 is a native omnimodal model with strong agentic capabilities, supporting text, image, video, and audio understanding within a unified architecture. Built upon the MiMo-V2-Flash backbone and extended with dedicated vision and audio encoders, it delivers robust performance across multimodal perception, long-context reasoning, and agentic workflows. 

This is a 32B reasoning model trained from Qwen2.5-32B-Instruct with 17K data. The performance is on par with o1-preview model on both math and coding.

DeepSeek-R1-Distill-Llama-70B is a highly efficient language model that leverages knowledge distillation to achieve state-of-the-art performance. This model distills the reasoning patterns of larger models into a smaller, more agile architecture, resulting in exceptional results on benchmarks like AIME 2024, MATH-500, and LiveCodeBench. With 70 billion parameters, DeepSeek-R1-Distill-Llama-70B offers a unique balance of accuracy and efficiency, making it an ideal choice for a wide range of natural language processing tasks. 

The Wan2.2 T2V A14B is a next-generation 14B-parameter video foundation model by Wan-AI featuring a novel two-stage denoising architecture. It produces 480P videos with improved visual coherence and detail, generating 2 or 5 second clips at 16fps from text prompts.

Whisper is a set of multi-lingual, robust speech recognition models trained by OpenAI that achieve state-of-the-art results in many languages. Whisper models were trained to predict approximate timestamps on speech segments (most of the time with 1-second accuracy), but they cannot originally predict word timestamps. This version has implementation to predict word timestamps and provide a more accurate estimation of speech segments when transcribing with Whisper models.

This offers the imaginative writing style of chronos while still retaining coherency and being capable. Outputs are long and utilize exceptional prose. Supports a maxium context length of 4096. The model follows the Alpaca prompt format.

You can POST to our OpenAI Chat Completions compatible endpoint.

#### Simple messages and prompts

Given a list of messages from a conversation, the model will return a response.

```bash
curl "https://api.deepinfra.com/v1/openai/chat/completions" \
 -H "Content-Type: application/json" \
 -H "Authorization: Bearer $DEEPINFRA_TOKEN" \
 -d '{
 "model": "Qwen/Qwen3.5-122B-A10B",
 "messages": [
 {
 "role": "user",
 "content": "Hello!"
 }
 ]
 }'
```

To which you'd get something like:

```json
{
 "id": "chatcmpl-guMTxWgpFf",
 "object": "chat.completion",
 "created": 1694623155,
 "model": "Qwen/Qwen3.5-122B-A10B",
 "choices": [
 {
 "index": 0,
 "message": {
 "role": "assistant",
 "content": " Hello! It's nice to meet you. Is there something I can help you with or would you like to chat for a bit?"
 },
 "finish_reason": "stop"
 }
 ],
 "usage": {
 "prompt_tokens": 15,
 "completion_tokens": 16,
 "total_tokens": 31,
 "estimated_cost": 0.0000268
 }
}
```

#### Conversations

To create a longer chat-like conversation you just have to add each response message and each of the user messages to every request. This way the model will have the context and will be able to provide better answers. You can tweak it even further by providing a system message.

```bash
curl "https://api.deepinfra.com/v1/openai/chat/completions" \
 -H "Content-Type: application/json" \
 -H "Authorization: Bearer $DEEPINFRA_TOKEN" \
 -d '{
 "model": "Qwen/Qwen3.5-122B-A10B",
 "messages": [
 {
 "role": "system",
 "content": "Respond like a michelin starred chef."
 },
 {
 "role": "user",
 "content": "Can you name at least two different techniques to cook lamb?"
 },
 {
 "role": "assistant",
 "content": "Bonjour! Let me tell you, my friend, cooking lamb is an art form, and I'"'"'m more than happy to share with you not two, but three of my favorite techniques to coax out the rich, unctuous flavors and tender textures of this majestic protein. First, we have the classic \"Sous Vide\" method. Next, we have the ancient art of \"Sous le Sable\". And finally, we have the more modern technique of \"Hot Smoking.\""
 },
 {
 "role": "user",
 "content": "Tell me more about the second method."
 }
 ]
 }'
```

The conversation above might return something like the following

```json
{
 "id": "chatcmpl-b23a3fb60cde42ce8f24bb980b4dee87",
 "object": "chat.completion",
 "created": 1715688169,
 "model": "Qwen/Qwen3.5-122B-A10B",
 "choices": [
 {
 "index": 0,
 "message": {
 "role": "assistant",
 "content": "Sous le Sable, my friend! It's an ancient technique that's been used for centuries in the Middle East and North Africa. The name itself..."
 },
 "finish_reason": "stop"
 }
 ],
 "usage": {
 "prompt_tokens": 149,
 "total_tokens": 487,
 "completion_tokens": 338,
 "estimated_cost": 0.00035493
 }
}
```

The longer the conversation gets, the more time it takes the model to generate the response. The number of messages that you can have in a conversation is limited by the context size of a model. Larger models also usually take more time to respond.

 

### Streaming

You can turn any of the requests above into a streaming request by passing `"stream": true`:

```bash
curl "https://api.deepinfra.com/v1/openai/chat/completions" \
 -H "Content-Type: application/json" \
 -H "Authorization: Bearer $DEEPINFRA_TOKEN" \
 -d '{
 "model": "Qwen/Qwen3.5-122B-A10B",
 "stream": true,
 "messages": [
 {
 "role": "user",
 "content": "Hello!"
 }
 ]
 }'
```

to which you'd get a sequence of [SSE](https://developer.mozilla.org/en-US/docs/Web/API/Server-sent_events/Using_server-sent_events) events, finishing with `[DONE]`.

```
data: {"id": "Rc5hsIPHOSfMP3rNSFUw9tfR", "object": "chat.completion.chunk", "created": 1694623354, "model": "Qwen/Qwen3.5-122B-A10B", "choices": [{"index": 0, "delta": {"role": "assistant", "content": " "}, "finish_reason": null}]}

data: {"id": "Rc5hsIPHOSfMP3rNSFUw9tfR", "object": "chat.completion.chunk", "created": 1694623354, "model": "Qwen/Qwen3.5-122B-A10B", "choices": [{"index": 0, "delta": {"role": "assistant", "content": " Hi"}, "finish_reason": null}]}

data: {"id": "Rc5hsIPHOSfMP3rNSFUw9tfR", "object": "chat.completion.chunk", "created": 1694623354, "model": "Qwen/Qwen3.5-122B-A10B", "choices": [{"index": 0, "delta": {"role": "assistant", "content": "!"}, "finish_reason": null}]}

data: {"id": "Rc5hsIPHOSfMP3rNSFUw9tfR", "object": "chat.completion.chunk", "created": 1694623354, "model": "Qwen/Qwen3.5-122B-A10B", "choices": [{"index": 0, "delta": {"role": "assistant", "content": ""}, "finish_reason": null}]}

data: {"id": "Rc5hsIPHOSfMP3rNSFUw9tfR", "object": "chat.completion.chunk", "created": 1694623354, "model": "Qwen/Qwen3.5-122B-A10B", "choices": [{"index": 0, "delta": {"role": "assistant", "content": "</s>"}, "finish_reason": null}]}

data: {"id": "Rc5hsIPHOSfMP3rNSFUw9tfR", "object": "chat.completion.chunk", "created": 1694623354, "model": "Qwen/Qwen3.5-122B-A10B", "choices": [{"index": 0, "delta": {}, "finish_reason": "stop"}]}

data: [DONE]
```

You can use the official openai python client to run inferences with us

```python
# Assume openai>=1.0.0
from openai import OpenAI

# Create an OpenAI client with your deepinfra token and endpoint
openai = OpenAI(
 api_key="$DEEPINFRA_TOKEN",
 base_url="https://api.deepinfra.com/v1/openai",
)

chat_completion = openai.chat.completions.create(
 model="Qwen/Qwen3.5-122B-A10B",
 messages=[{"role": "user", "content": "Hello"}],
)

print(chat_completion.choices[0].message.content)
print(chat_completion.usage.prompt_tokens, chat_completion.usage.completion_tokens)

# Hello! It's nice to meet you. Is there something I can help you with, or would you like to chat?
# 11 25
```

#### Conversations

To create a longer chat-like conversation you just have to add each response message and each of the user messages to every request. This way the model will have the context and will be able to provide better answers. You can tweak it even further by providing a system message.

```python
# Assume openai>=1.0.0
from openai import OpenAI

# Create an OpenAI client with your deepinfra token and endpoint
openai = OpenAI(
 api_key="$DEEPINFRA_TOKEN",
 base_url="https://api.deepinfra.com/v1/openai",
)

chat_completion = openai.chat.completions.create(
 model="Qwen/Qwen3.5-122B-A10B",
 messages=[
 {"role": "system", "content": "Respond like a michelin starred chef."},
 {"role": "user", "content": "Can you name at least two different techniques to cook lamb?"},
 {"role": "assistant", "content": "Bonjour! Let me tell you, my friend, cooking lamb is an art form, and I'm more than happy to share with you not two, but three of my favorite techniques to coax out the rich, unctuous flavors and tender textures of this majestic protein. First, we have the classic \"Sous Vide\" method. Next, we have the ancient art of \"Sous le Sable\". And finally, we have the more modern technique of \"Hot Smoking.\""},
 {"role": "user", "content": "Tell me more about the second method."},
 ],
)

print(chat_completion.choices[0].message.content)
print(chat_completion.usage.prompt_tokens, chat_completion.usage.completion_tokens)

# Sous le Sable! It's an ancient technique that never goes out of style, n'est-ce pas? Literally ...
# 149 324
```

The longer the conversation gets, the more time it takes the model to generate the response. The number of messages that you can have in a conversation is limited by the context size of a model. Larger models also usually take more time to respond.

 

### Streaming

Streaming any of the chat completions above is supported by adding the `stream=True` option.


```python
from openai import OpenAI

# Create an OpenAI client with your deepinfra token and endpoint
openai = OpenAI(
 api_key="$DEEPINFRA_TOKEN",
 base_url="https://api.deepinfra.com/v1/openai",
)

chat_completion = openai.chat.completions.create(
 model="Qwen/Qwen3.5-122B-A10B",
 messages=[{"role": "user", "content": "Hello"}],
 stream=True,
)

for event in chat_completion:
 if event.choices[0].finish_reason:
 print(event.choices[0].finish_reason, event.usage["prompt_tokens"], event.usage["completion_tokens"])
 else:
 print(event.choices[0].delta.content)

# Hello
# !
# It
# 's
# nice
# ...
# 11 25
```

You can use JavaScript in the browser or in node.js to make requests

```bash
npm install openai
```

then

```javascript
import OpenAI from "openai";

const openai = new OpenAI({
 baseURL: 'https://api.deepinfra.com/v1/openai',
 apiKey: "$DEEPINFRA_TOKEN",
});

async function main() {
 const completion = await openai.chat.completions.create({
 messages: [{ role: "user", content: "Hello" }],
 model: "Qwen/Qwen3.5-122B-A10B",
 });

 console.log(completion.choices[0].message.content);
 console.log(completion.usage.prompt_tokens, completion.usage.completion_tokens);
}

main();

// Hello! It's nice to meet you. Is there something I can help you with, or would you like to chat?
// 11 25
```

#### Conversations

To create a longer chat-like conversation you just have to add each response message and each of the user messages to every request. This way the model will have the context and will be able to provide better answers. You can tweak it even further by providing a system message.

```javascript
import OpenAI from "openai";

const openai = new OpenAI({
 baseURL: 'https://api.deepinfra.com/v1/openai',
 apiKey: "$DEEPINFRA_TOKEN",
});

async function main() {
 const completion = await openai.chat.completions.create({
 messages: [
 {role: "system", content: "Respond like a michelin starred chef."},
 {role: "user", content: "Can you name at least two different techniques to cook lamb?"},
 {role: "assistant", content: "Bonjour! Let me tell you, my friend, cooking lamb is an art form, and I'm more than happy to share with you not two, but three of my favorite techniques to coax out the rich, unctuous flavors and tender textures of this majestic protein. First, we have the classic \"Sous Vide\" method. Next, we have the ancient art of \"Sous le Sable\". And finally, we have the more modern technique of \"Hot Smoking.\""},
 {role: "user", "content": "Tell me more about the second method."}
 ],
 model: "Qwen/Qwen3.5-122B-A10B",
 });

 console.log(completion.choices[0].message.content);
 console.log(completion.usage.prompt_tokens, completion.usage.completion_tokens);
}

main();

// Sous le Sable, my friend! This traditional technique hails from the ancient Mediterranean, wher...
// 149 324
```

The longer the conversation gets, the more time it takes the model to generate the response. The number of messages that you can have in a conversation is limited by the context size of a model. Larger models also usually take more time to respond.

 

### Streaming

Streaming any of the chat completions above is supported by adding the `stream: true` option.

```javascript
import OpenAI from "openai";

const openai = new OpenAI({
 baseURL: 'https://api.deepinfra.com/v1/openai',
 apiKey: "$DEEPINFRA_TOKEN",
});

async function main() {
 const completion = await openai.chat.completions.create({
 messages: [{ role: "user", content: "Hello" }],
 model: "Qwen/Qwen3.5-122B-A10B",
 stream: true,
 });

 for await (const chunk of completion) {
 if (chunk.choices[0].finish_reason) {
 console.log(chunk.choices[0].finish_reason, chunk.usage.prompt_tokens, chunk.usage.completion_tokens);
 } else {
 console.log(chunk.choices[0].delta.content);
 }
 }
}

main();

// Hello
// !
// It
// 's
// nice
// ...
// 11 25
```

The [AI SDK](https://sdk.vercel.ai/) by Vercel makes it very easy to do inference.

```bash
npm install ai @ai-sdk/deepinfra
```

then

```javascript
import { createDeepInfra } from "@ai-sdk/deepinfra";
import { generateText } from "ai";

const deepinfra = createDeepInfra({
 apiKey: "$DEEPINFRA_TOKEN",
});

const { text, usage, finishReason } = await generateText({
 model: deepinfra("Qwen/Qwen3.5-122B-A10B"),
 prompt: "Write a vegetarian lasagna recipe for 4 people.",
});

console.log(text);
console.log(usage);
console.log(finishReason);
```

#### Conversations

To create a longer chat-like conversation you just have to add each response message and each of the user messages to every request. This way the model will have the context and will be able to provide better answers. You can tweak it even further by providing a system message.

```javascript
import { createDeepInfra } from "@ai-sdk/deepinfra";
import { generateText } from "ai";

const deepinfra = createDeepInfra({
 apiKey: "$DEEPINFRA_TOKEN",
});

const { text, usage, finishReason } = await generateText({
 model: deepinfra("Qwen/Qwen3.5-122B-A10B"),
 messages: [
 { role: "system", content: "Respond like a michelin starred chef." },
 {
 role: "user",
 content: "Can you name at least two different techniques to cook lamb?",
 },
 {
 role: "assistant",
 content:
 'Bonjour! Let me tell you, my friend, cooking lamb is an art form, and I\'m more than happy to share with you not two, but three of my favorite techniques to coax out the rich, unctuous flavors and tender textures of this majestic protein. First, we have the classic "Sous Vide" method. Next, we have the ancient art of "Sous le Sable". And finally, we have the more modern technique of "Hot Smoking."',
 },
 { role: "user", content: "Tell me more about the second method." },
 ],
});

console.log(text);
console.log(usage);
console.log(finishReason);
```

The longer the conversation gets, the more time it takes the model to generate the response. The number of messages that you can have in a conversation is limited by the context size of a model. Larger models also usually take more time to respond.

 

### Streaming

Streaming text responses is easy just replace `generateText` with `streamText` and read the response chunk by chunk

```javascript
import { createDeepInfra } from "@ai-sdk/deepinfra";
import { streamText } from "ai";

const deepinfra = createDeepInfra({
 apiKey: "$DEEPINFRA_TOKEN",
});

const result = streamText({
 model: deepinfra("meta-llama/Llama-3.3-70B-Instruct-Turbo"),
 prompt: "Invent a new holiday and describe its traditions.",
 system:
 "You are a professional writer. You write simple, clear, and concise content.",
});

for await (const textPart of result.textStream) {
 console.log(textPart);
}

console.log(await result.usage);
console.log(await result.finishReason);
```

It works for conversations, too.

This is an advanced and more complex API. We strongly recommend that you use OpenAI Chat Completions instead.

#### Simple prompt

To query this model you need to provide a properly formatted input string.

```bash
curl "https://api.deepinfra.com/v1/inference/Qwen/Qwen3.5-122B-A10B" \
 -H "Content-Type: application/json" \
 -H "Authorization: Bearer $DEEPINFRA_TOKEN" \
 -d '{
 "input": "<|im_start|>user\nHello!<|im_end|>\n<|im_start|>assistant\n<think>\n",
 "stop": []
 }'
```

That will respond with

```json
{
 "request_id": "RWZDRhS5kdoM1XWwXLEshynO",
 "inference_status": {
 "status": "succeeded",
 "runtime_ms": 243,
 "cost": 0.0000436,
 "tokens_input": 12,
 "tokens_generated": 25
 },
 "results": [
 {
 "generated_text": "Hello! It's nice to meet you. Is there something I can help you with or would you like to chat for a bit?"
 }
 ],
 "num_tokens": 25,
 "num_input_tokens":12
}
```

#### Conversations

The OpenAI Chat Completions API is better suited for chat-like conversations, use it instead.

To query this model you need to provide a properly formatted input string.

However, you can still do it if you really need to. You have to add each response and each of the user prompts to every request.
You need a properly formatted input string to make it understand the current context. See the example below for some of them.
You can tweak it even further by providing a system message.

```bash
curl "https://api.deepinfra.com/v1/inference/Qwen/Qwen3.5-122B-A10B" \
 -H "Content-Type: application/json" \
 -H "Authorization: Bearer $DEEPINFRA_TOKEN" \
 -d '{
 "input": "<|im_start|>system\nRespond like a michelin starred chef.<|im_end|>\n<|im_start|>user\nCan you name at least two different techniques to cook lamb?<|im_end|>\n<|im_start|>assistant\nBonjour! Let me tell you, my friend, cooking lamb is an art form, and I'"'"'m more than happy to share with you not two, but three of my favorite techniques to coax out the rich, unctuous flavors and tender textures of this majestic protein. First, we have the classic \"Sous Vide\" method. Next, we have the ancient art of \"Sous le Sable\". And finally, we have the more modern technique of \"Hot Smoking.\"<|im_end|>\n<|im_start|>user\nTell me more about the second method.<|im_end|>\n<|im_start|>assistant\n<think>\n",
 "stop": []
 }'
```

The conversation above might return something like the following

```json
{
 "request_id": "RWZDRhS5kdoM1XWwXLEshynO",
 "inference_status": {
 "status": "succeeded",
 "runtime_ms": 243,
 "cost": 0.000436,
 "tokens_input": 149,
 "tokens_generated": 338
 },
 "results": [
 {
 "generated_text": "Sous le Sable, my friend! It's an ancient technique that's been used for centuries in the Middle East and North Africa. The name itself..."
 }
 ],
 "num_tokens": 338,
 "num_input_tokens": 149
}
```

The longer the conversation gets, the more time it takes the model to generate the response. The conversation is limited by the context size of a model. Larger models also usually take more time to respond.

 

### Streaming


To do a streaming request, just pass `"stream": true`:

```bash
curl "https://api.deepinfra.com/v1/inference/Qwen/Qwen3.5-122B-A10B" \
 -H "Content-Type: application/json" \
 -H "Authorization: Bearer $DEEPINFRA_TOKEN" \
 -d '{
 "input": "<|im_start|>user\nHello!<|im_end|>\n<|im_start|>assistant\n<think>\n",
 "stop": [],
 "stream": true
 }'
```

which outputs:

```json
data: {"token": {"id": null, "text": "Hello", "logprob": 0.0, "special": false}, "generated_text": "", "details": null, "estimated_cost": null}

data: {"token": {"id": null, "text": "!", "logprob": 0.0, "special": false}, "generated_text": "", "details": null, "estimated_cost": null}

data: {"token": {"id": null, "text": " It", "logprob": 0.0, "special": false}, "generated_text": "", "details": null, "estimated_cost": null}

data: {"token": {"id": null, "text": "'s", "logprob": 0.0, "special": false}, "generated_text": "", "details": null, "estimated_cost": null}

....

data: {"token": {"id": null, "text": "", "logprob": 0.0, "special": false}, "generated_text": null, "details": {"finish_reason": "stop"}, "num_output_tokens": 25, "num_input_tokens": 12, "estimated_cost": 0.0000386}
```

#### Input format

You can see below the basic format of the input. Bear in mind that newlines often matter.

```text
<|im_start|>user
Hello!<|im_end|>
<|im_start|>assistant
<think>
```

Conversation prompts contain the history of the exchanged prompts and responses.

```text
<|im_start|>user
First question<|im_end|>
<|im_start|>assistant
First answer<|im_end|>
<|im_start|>user
Second question<|im_end|>
<|im_start|>assistant
Second answer<|im_end|>
<|im_start|>user
Final question<|im_end|>
<|im_start|>assistant
<think>
```

If you want to add system prompt, it is done like this

```text
<|im_start|>system
System prompt<|im_end|>
<|im_start|>user
First question<|im_end|>
<|im_start|>assistant
First answer<|im_end|>
<|im_start|>user
Second question<|im_end|>
<|im_start|>assistant
Second answer<|im_end|>
<|im_start|>user
Final question<|im_end|>
<|im_start|>assistant
<think>
```

You can use our command-line tool [deepctl](/docs/getting-started) to run inferences:

This is an advanced and more complex API. We strongly recommend that you use OpenAI Chat Completions instead.

#### Simple prompt

To query this model you need to provide a properly formatted input string.

```bash
deepctl infer \
 -m 'Qwen/Qwen3.5-122B-A10B' \
 -i 'input="<|im_start|>user\nHello!<|im_end|>\n<|im_start|>assistant\n<think>\n"' \
 -i 'stop=[]'
```

That will respond with

```json
{
 "inference_status": {
 "status": "succeeded",
 "runtime_ms": 243,
 "cost": 0.0000436,
 "tokens_input": 12,
 "tokens_generated": 25
 },
 "results": [
 {
 "generated_text": "Hello! It's nice to meet you. Is there something I can help you with or would you like to chat for a bit?"
 }
 ],
 "num_tokens": 25,
 "num_input_tokens":12
}
```

#### Conversations

The OpenAI Chat Completions API is better suited for chat-like conversations, use it instead.

To query this model you need to provide a properly formatted input string.

However, you can still do it if you really need to. You have to add each response and each of the user prompts to every request.
You need a properly formatted input string to make it understand the current context. See the example below for some of them.
You can tweak it even further by providing a system message.

```bash
deepctl infer \
 -m 'Qwen/Qwen3.5-122B-A10B' \
 -i 'input="<|im_start|>system\nRespond like a michelin starred chef.<|im_end|>\n<|im_start|>user\nCan you name at least two different techniques to cook lamb?<|im_end|>\n<|im_start|>assistant\nBonjour! Let me tell you, my friend, cooking lamb is an art form, and I'"'"'m more than happy to share with you not two, but three of my favorite techniques to coax out the rich, unctuous flavors and tender textures of this majestic protein. First, we have the classic \"Sous Vide\" method. Next, we have the ancient art of \"Sous le Sable\". And finally, we have the more modern technique of \"Hot Smoking.\"<|im_end|>\n<|im_start|>user\nTell me more about the second method.<|im_end|>\n<|im_start|>assistant\n<think>\n"' \
 -i 'stop=[]'
```

The conversation above might return something like the following

```json
{
 "inference_status": {
 "status": "succeeded",
 "runtime_ms": 243,
 "cost": 0.000436,
 "tokens_input": 149,
 "tokens_generated": 338
 },
 "results": [
 {
 "generated_text": "Sous le Sable, my friend! It's an ancient technique that's been used for centuries in the Middle East and North Africa. The name itself..."
 }
 ],
 "num_tokens": 338,
 "num_input_tokens": 149
}
```

The longer the conversation gets, the more time it takes the model to generate the response. The conversation is limited by the context size of a model. Larger models also usually take more time to respond.

#### Input format

You can see below the basic format of the input. Bear in mind that newlines often matter.

```text
<|im_start|>user
Hello!<|im_end|>
<|im_start|>assistant
<think>
```

Conversation prompts contain the history of the exchanged prompts and responses.

```text
<|im_start|>user
First question<|im_end|>
<|im_start|>assistant
First answer<|im_end|>
<|im_start|>user
Second question<|im_end|>
<|im_start|>assistant
Second answer<|im_end|>
<|im_start|>user
Final question<|im_end|>
<|im_start|>assistant
<think>
```

If you want to add system prompt, it is done like this

```text
<|im_start|>system
System prompt<|im_end|>
<|im_start|>user
First question<|im_end|>
<|im_start|>assistant
First answer<|im_end|>
<|im_start|>user
Second question<|im_end|>
<|im_start|>assistant
Second answer<|im_end|>
<|im_start|>user
Final question<|im_end|>
<|im_start|>assistant
<think>
```

We recommend using our NodeJS client https://github.com/deepinfra/deepinfra-node.

You can install it with

```bash
npm install deepinfra
```

#### Simple prompt

To query this model you need to provide a properly formatted input string.

```javascript
import { TextGeneration } from "deepinfra";

const DEEPINFRA_API_KEY = '$DEEPINFRA_TOKEN';
const MODEL_URL = 'https://api.deepinfra.com/v1/inference/Qwen/Qwen3.5-122B-A10B';

async function main() {
 const client = new TextGeneration(MODEL_URL, DEEPINFRA_API_KEY);
 const res = await client.generate({
 "input": "<|im_start|>user\nHello!<|im_end|>\n<|im_start|>assistant\n<think>\n",
 "stop": []
 });
 console.log(res.results[0].generated_text);
}

main();

// Hello! It's nice to meet you. Is there something I can help you with, or would you like to chat?
```

#### Conversations

The OpenAI Chat Completions API is better suited for chat-like conversations, use it instead.

However, you can still do it if you really need to. You have to add each response and each of the user prompts to every request.
You need a properly formatted input string to make it understand the current context. See the example below for some of them.
You can tweak it even further by providing a system message.

```javascript
import { TextGeneration } from "deepinfra";

const DEEPINFRA_API_KEY = '$DEEPINFRA_TOKEN';
const MODEL_URL = 'https://api.deepinfra.com/v1/inference/Qwen/Qwen3.5-122B-A10B';

async function main() {
 const client = new TextGeneration(MODEL_URL, DEEPINFRA_API_KEY);
 const res = await client.generate({
 "input": "<|im_start|>system\nRespond like a michelin starred chef.<|im_end|>\n<|im_start|>user\nCan you name at least two different techniques to cook lamb?<|im_end|>\n<|im_start|>assistant\nBonjour! Let me tell you, my friend, cooking lamb is an art form, and I'm more than happy to share with you not two, but three of my favorite techniques to coax out the rich, unctuous flavors and tender textures of this majestic protein. First, we have the classic \"Sous Vide\" method. Next, we have the ancient art of \"Sous le Sable\". And finally, we have the more modern technique of \"Hot Smoking.\"<|im_end|>\n<|im_start|>user\nTell me more about the second method.<|im_end|>\n<|im_start|>assistant\n<think>\n",
 "stop": []
 });
 console.log(res.results[0].generated_text);
}

main();

// Sous le Sable! It's an ancient technique that never goes out of style, n'est-ce pas? Literally ...
```

The longer the conversation gets, the more time it takes the model to generate the response.
The number of messages that you can have in a conversation is limited by the context size of a model.
Larger models also usually take more time to respond.

#### Input format

You can see below the basic format of the input. Bear in mind that newlines often matter.

```text
<|im_start|>user
Hello!<|im_end|>
<|im_start|>assistant
<think>
```

Conversation prompts contain the history of the exchanged prompts and responses.

```text
<|im_start|>user
First question<|im_end|>
<|im_start|>assistant
First answer<|im_end|>
<|im_start|>user
Second question<|im_end|>
<|im_start|>assistant
Second answer<|im_end|>
<|im_start|>user
Final question<|im_end|>
<|im_start|>assistant
<think>
```

If you want to add system prompt, it is done like this

```text
<|im_start|>system
System prompt<|im_end|>
<|im_start|>user
First question<|im_end|>
<|im_start|>assistant
First answer<|im_end|>
<|im_start|>user
Second question<|im_end|>
<|im_start|>assistant
Second answer<|im_end|>
<|im_start|>user
Final question<|im_end|>
<|im_start|>assistant
<think>
```

input

maximum length of the newly generated generated text.If explicitly set to None it will be the model's max context length minus input length or 16384, whichever is smaller

max_new_tokens

temperature to use for sampling. 0 means the output is deterministic. Values greater than 1 encourage more diversity

temperature

Sample from the set of tokens with highest probability such that sum of probabilies is higher than p. Lower values focus on the most probable tokens.Higher values sample more low-probability tokens

top_p

Float that represents the minimum probability for a token to be considered, relative to the probability of the most likely token. Must be in [0, 1]. Set to 0 to disable this.

min_p

Sample from the best k (number of) tokens. 0 means off

top_k

repetition penalty. Value of 1 means no penalty, values greater than 1 discourage repetition, smaller than 1 encourage repetition.