OpenRouterBYOK
About OpenRouter
- Base URL
Models
Aion 3.0 Mini is Aion Labs' reasoning model built on the DeepSeek family. Several specialized models contribute to each response, giving stronger narrative structure and a more nuanced treatment of darker themes, so it suits fiction. Context is 131,072 tokens.
Aion 3.5 Mini is Aion Labs' compact reasoning model built on the GLM family. Several specialized models contribute to each response, which suits narrative writing. It has a 262,144-token context window and takes text input.
Aion 3.5 is Aion Labs' reasoning model built on the GLM family of models. Several specialized models contribute to each response, giving stronger narrative structure and a more nuanced treatment of mature and darker themes, so it suits fiction writing. Context is 262,144 tokens.
Claude Fable 5.1 is Anthropic's successor to Claude Fable 5 for long-running agentic coding, knowledge work and research. It has a 1M-token context window, 128K max output, tool use, and text, image and file input.
Claude Fable 5.2 is Anthropic's frontier model for the most demanding technical work, with gains over Fable 5.1 in coding, reasoning and vision. It supports text and image input, a 1M-token context window, 128K max output and always-on adaptive thinking.
Claude Fable 5 is Anthropic's Mythos-class model for autonomous knowledge work and coding. It has a 1M-token context window, 128K max output, tool use, and text, image and file input.
Claude Haiku 4.5 is Anthropic's fast Claude model for coding, with more than twice the speed of larger Claude models. It has 64,000 max output tokens, tool use, and text, image and file input.
Claude Haiku 5.5 is Anthropic's fastest model in the Claude 5.x lineup, built for rapid ideation, summarization and classification. It has vision input, tool use and always-on adaptive thinking.
Claude Opus 4.1 is Anthropic's model for complex analysis, advanced coding and multi-step reasoning. It supports tool use and streaming, with a 200,000-token context window and text, image and file input.
Claude Opus 4.5 is Anthropic's frontier-class model for complex analysis, advanced coding and multi-step reasoning. It supports tool use, 64,000 max output tokens, and text, image and file input.
Claude Opus 4.6 is Anthropic's flagship-tier Claude 4 model for reasoning, advanced coding and nuanced analysis. It has 128K max output, tool use, and text, image and file input.
Claude Opus 4.7 is Anthropic's model for sophisticated research, advanced coding and multi-step agentic workflows. It has 128K max output, tool use, and text, image and file input.
Claude Opus 4.8 is Anthropic's most advanced Claude Opus model, built for complex reasoning, frontier-level coding and in-depth research. It has 128K max output and text, image and file input.
Claude Opus 5.5 is Anthropic's successor to Claude Opus 5 for long-running agentic coding and knowledge work. It has a 1M-token context window, vision input, tool use and always-on adaptive thinking, suited to multi-step changes in large codebases.
Claude Opus 5 Fast is the fast-mode variant of Anthropic's Claude Opus 5, tuned for higher output speed. It has the same capabilities and a 1,000,000-token context window, with tool use and text, image and file input.
Claude Opus 5 is Anthropic's flagship model for demanding reasoning, coding and long-horizon agentic work. It handles end-to-end software tasks, code review, bug finding and chart analysis.
Claude Sonnet 4.5 is Anthropic's model that balances speed and intelligence for demanding text, coding and research tasks. It has 64,000 max output tokens, tool use, and text, image and file input.
Claude Sonnet 4.6 is Anthropic's balanced model for coding, analysis and reasoning. It supports tool use, up to 64,000 output tokens, and text, image and file input.
Claude Sonnet 5.5 is Anthropic's successor to Claude Sonnet 5 for well-scoped everyday work such as building features, fixing bugs, drafting and review. It has a 1M-token context window, 128K max output, vision input, tool use and adaptive thinking on by default.
Claude Sonnet 5 is Anthropic's agentic model with strong coding, tool use and reasoning. It has a 1M-token context window, 64,000 max output tokens, and takes text, image and file input.
Trinity Large Thinking is Arcee AI's open-weight reasoning model. It is a 398B sparse mixture-of-experts with about 13B active parameters per token, built for tool calling, multi-step planning and agent workflows. It writes its reasoning in think blocks before the answer.
ERNIE 5.5 Flash Preview is Baidu's mixture-of-experts reasoning model with a 1M-token context window. It has 280B total and 14B active parameters, with tool calling, structured outputs and reasoning effort control. It targets coding, agentic tasks and deep research.
Seed 2.1 Turbo is ByteDance Seed's multimodal model for coding and long-horizon agent workflows. It takes text, image and video input, supports tool use, and has a 262,144-token context window.
Seed 2.0 Code is a ByteDance Seed model optimized for agentic coding, frontend development and multilingual programming. It takes text, image and video input, and supports tool use.
Command A+ is Cohere's first Mixture-of-Experts model, with vision input, reasoning, and agentic tool use. It has 25B active parameters, supports 48 languages, and offers a 128K-token context window with up to 64K tokens of output. It is released open-weight.
Command A is Cohere's flagship model for agentic, multilingual, and enterprise use cases. It is optimized for tool use, retrieval-augmented generation, and complex reasoning, supports 23 languages, and offers a 256K-token context window. Released open-weight for research use.
North Mini Code is Cohere's first agentic coding model, a sparse mixture-of-experts with 30B total and 3B active parameters. It has a 256K-token context window with up to 64K tokens of output and interleaved reasoning with tool use. It is open-weight under the Apache 2.0 license.
DeepSeek-R1 is DeepSeek's open-weight reasoning model, a 671B mixture-of-experts with 37B active parameters and a 131,072-token context. It is trained with large-scale reinforcement learning and released under the MIT license. It targets math, code and reasoning.
DeepSeek V3.2 Exp is an experimental DeepSeek model built on V3.1-Terminus that adds DeepSeek Sparse Attention to improve long-context training and inference efficiency. It matches V3.1-Terminus on public benchmarks, and is released with open weights under the MIT license.
DeepSeek-V4-Flash is DeepSeek's open-weight fast mixture-of-experts model for high-throughput chat and coding. It accepts text and image input and returns text. It supports reasoning, tool calling and agentic use.
DeepSeek V4 Pro is DeepSeek's open-weight high-capability reasoning model. It is a text model for chat and coding with streaming and function calling.
DeepSeek V4.1 Flash 1017 is the October 17, 2026 snapshot of DeepSeek V4.1 Flash. It is an open-weight sparse mixture-of-experts model with native vision input, text output and a 1M-token context window.
DeepSeek V4.1 Flash is the open-weight, cost-efficient sparse mixture-of-experts tier of the V4.1 family. It has 552B total parameters with 8B active on input and 16B on output, 1M context, and native vision in with text out. This is the original September 2026 checkpoint.
Dots3-Note Preview is Dots Studio's open-weight mixture-of-experts reasoning model, released as a preview. It has 280B total and 16B active parameters with a 512,000-token context window. It takes text and image input and supports function calling.
Glimpse is Elorian's visual reasoning model for image and video understanding. It tracks objects across motion and viewpoints and reasons about spatial relationships and physical cause and effect. It takes text, image, and video input and answers in text.
Ember-1 is a reasoning model from Fireworks Research built on Kimi K3. It is tuned to produce shorter reasoning traces, with tool use, image input, and a 1M-token context window.
Gemini 2.5 Flash-Lite is Google's lightest Gemini 2.5 Flash variant, built for high-throughput inference. It has a 1M-token context window and takes text, image, video, and audio input. It supports tool calling and coding.
Gemini 2.5 Flash is Google's fast Gemini model with strong reasoning, designed for high-throughput tasks. It takes text and image input and generates text. It supports tool calling and coding.
Gemini 2.5 Pro is Google's advanced Gemini model for deep reasoning over long context. It accepts text, image, video, and audio input and generates text. It fits complex research, coding, and analytical workflows.
Gemini 3 Flash is Google's fast Gemini model with strong reasoning, coding, and multimodal understanding. It accepts text, image, video, and audio input. It supports tool calling.
Gemini 3.1 Flash Lite is Google's lightest Gemini model, built for high-throughput tasks. It takes text and image input and generates text. It supports tool calling, coding, and streaming responses.
Gemini 3.1 Pro is Google's most intelligent Gemini model, with improved reasoning and a medium thinking level. It takes text and image input. It supports tool calling and coding.
Gemini 3.5 Flash Lite is Google's high-throughput multimodal model for agentic workflows, complex coding, and long-horizon tasks. It has a 1M-token context window and accepts text and image input. It supports tool calling.
Gemini 3.5 Flash is Google's multimodal model for agentic workflows, complex coding, and long-horizon tasks. It takes text and image input and generates text. It supports tool calling and streaming.
Gemini 3.6 Flash is Google's frontier-level multimodal model that pairs fast inference with strong reasoning and coding. It accepts text and image input. It supports tool calling.
Gemini 3.7 Flash is Google's multimodal model for fast agentic workflows, coding, and complex multi-step reasoning. It accepts text, image, video, file, and audio input.
Gemini 3.8 Flash is Google's multimodal model for fast agentic workflows, coding, and complex multi-step reasoning. It has a 1,048,576-token context window. It accepts text, image, video, file, and audio input.
Gemini 4 Argon is Google's frontier model for long-horizon software engineering, enterprise knowledge work, and cybersecurity defense. It accepts text, image, and video input.
Gemini 4 Carbon is Google's strongest model, built for long-horizon agentic work such as multi-step coding and debugging. It reads text, images, and video and has a 1,048,576-token context window. It also targets knowledge work, research, and cybersecurity.
Gemini 4 Hydrogen is the speed and efficiency model of Google's Gemini 4 family, built for everyday coding and agentic workflows. It reads text, images, video, audio, and PDFs, answers in text, and has a 1,048,576-token context window with adjustable thinking levels.
Gemini Decision 1 is Google's decision model for the Gemini Decisions API. It scores yes/no, choice, and rubric questions on one text, image, audio, video, or PDF input in a single pass, with a 262,144-token context window.
Gemini Embedding 2 is Google's first multimodal embedding model, mapping text, images, video, audio and PDFs into one 3,072-dimension space. This listing takes text input only, with an 8,192-token window.
Gemma 3 4B IT is Google's instruction-tuned 4B open model with text and image input. It has a 128K context window and multilingual support in over 140 languages. It supports tool calling.
Gemma 4 26B A4B is an open-weight model from Google's Gemma 4 family, built from Gemini 3 research to maximize intelligence per parameter. It takes text and image input and generates text. It supports reasoning, tool calling, and coding.
Gemma 4 31B is Google's open-weight dense open reasoning model for complex reasoning, agentic workflows, coding, and multimodal understanding. It takes text and image input and generates text. It supports tool calling and a thinking mode.
Granite 4.0 H Micro is IBM's open-weight instruct model in the Granite 4.0 family, built for instruction following and function calling. It has a 131,000-token context window. It suits retrieval-augmented generation, multi-agent workflows and edge deployments.
Granite 4.2 8B is the open-weight, mid-size dense reasoning model in IBM's Granite 4.2 family. It offers full, low-effort and non-thinking modes plus tool calling. Context window is 131,072 tokens, with up to 117,964 output tokens.
Mercury 2.5 is Inception Labs' diffusion language model for fast text generation and reasoning. It has a 260,000-token context window and supports tool calling and a thinking mode.
Mercury 3 is Inception's diffusion language model, the lab's most capable to date, built for agentic workflows and real-time applications. It supports tunable reasoning, tool calling, and structured output. The context window is 1M tokens.
Mercury Decide is Inception's structured decision model. It takes application state and typed questions and returns a choice, a score, or a yes/no answer, each with a calibrated probability. No output tokens are generated, and it answers up to 14 decisions per second.
Ling-3.0-flash-Fin is inclusionAI's open-weight finance-focused mixture-of-experts model, built on Ling-3.0-flash with 124B total and 5.1B active parameters. It targets investment research, market analysis and financial reasoning. It keeps general reasoning and coding skills.
Ling-3.0-flash-Sante is inclusionAI's open-weight health and medicine mixture-of-experts model, built on Ling-3.0-flash with 124B total and 5.1B active parameters. It targets medical knowledge reasoning, clinical safety and evidence-based retrieval. It keeps coding skills.
Ling-3.0-flash-VL is inclusionAI's open-weight native multimodal mixture-of-experts model with 124B parameters and about 5.5B active per token. It adds image and video understanding. It suits visual reasoning, document and chart reading, and agentic GUI tasks.
Ling-3.0-flash is inclusionAI's open-weight instruct mixture-of-experts model with 124B parameters and about 5.1B active per token. It is built for token efficiency and production-scale agentic inference. It suits coding, document processing and lightweight agent workflows.
Ling-3.1-flash is inclusionAI's open-weight hybrid reasoning mixture-of-experts model with 25B active parameters out of 560B total. It is a text model with streaming, function calling, coding and agentic use.
Schematron V2 Small is inference.net's 3B extraction model that turns messy HTML into clean JSON matching a schema you supply. It is the family's highest-quality tier, with a 128K-token context and strict JSON schema output.
Schematron V2 Turbo is inference.net's 3B extraction model that turns messy HTML into clean JSON matching a schema you supply. It is the family's throughput-optimized option, faster than V2 Small, with a 128K-token context and strict JSON schema output.
Kev 4B is Jared Palmer's small open-weight decision model, a LoRA adapter and pointer head on Qwen3.5-4B-Base. It returns a calibrated probability per typed question in one forward pass. It does not generate text.
Mi:dm K 3.0 Pro is KT's Mixture-of-Experts reasoning model, with 120B total and 12B active parameters. It mixes Mamba, MoE and attention layers, takes text in and gives text out, and has a 261K context window. KT serves it through its Token Factory API.
LFM 2.5 2.6B is Liquid AI's compact 2.6B-parameter chat model for general assistant and coding tasks. It allows up to 32,768 output tokens. It takes text in and streams text out.
LFM 2.5 24B A2B is Liquid AI's open-weight reasoning mixture-of-experts model with 24B total and 2.45B active parameters. It always reasons before it answers and supports tool use and coding. It has a 131,072-token context window and uses the LFM Open License v1.0.
LongCat-2.0 is Meituan's open-weight sparse mixture-of-experts language model with 1.6T total and 48B active parameters. It targets coding, repository-level changes, and agentic workflows. It supports streaming, function calling, and structured output.
Llama 3.1 70B Instruct is Meta's instruction-tuned model for chat, reasoning, and multilingual use. It supports coding, function calling, and streaming, with text in and text out.
Llama 3.1 8B Instruct is Meta's compact instruction-tuned model, optimized for fast inference and edge deployments. It supports chat, reasoning, coding, function calling, and streaming, with text in and text out.
Llama 3.3 70B Instruct is Meta's open-weight instruction-tuned model with 70B parameters and a 131072-token context window. It is tuned for multilingual dialogue, with tool use on par with Llama 3.1 70B. It suits chat, reasoning, and coding.
Llama 4 Scout 17B 16E is Meta's open-weight multimodal mixture-of-experts model with 17B active and 109B total parameters. It takes text and image input and has a 10M-token input context. It fits on a single H100 GPU.
Llama Guard 4 12B is Meta's natively multimodal safety classifier with 12B parameters. It labels LLM prompts and responses as safe or unsafe and lists violated categories from the MLCommons hazards taxonomy. Text and image go in, text comes out.
Muse Ember is Meta's smallest and most efficient model in the Muse family, built for reasoning and everyday agent work. It has a 1M-token context window with up to 524,288 output tokens. It takes text, image, video, file and audio input and returns text.
Muse Glimmer 30B is Meta's dense multimodal instruct model with about 30B parameters, distilled from Muse Spark. It takes text and image input and returns text, with reasoning and function calling. Weights are open under Apache 2.0.
Muse Spark 1.1 is Meta's original multimodal reasoning model in the Muse Spark family, built for agentic tasks. It takes text, image, video, audio and file input, returns text, and has a 1M-token context window. It supports function calling and structured output.
Muse Spark 1.2 Contributor is the Contributor-mode tier of Meta's Muse Spark 1.2 reasoning model, for experimentation and early agentic work. Meta may use your prompts, files, and responses to train future models, so avoid confidential work.
Muse Spark 1.2 is Meta's multimodal reasoning model for complex agentic tasks. It accepts text, images, video, audio, and PDF documents, returns text, and offers a 1M-token context window.
Muse Spark 1.3 Contributor is the Contributor-mode tier of Meta's Muse Spark 1.3 reasoning model, for experimentation and early agentic work. Meta may use your prompts, files, and responses to train future models, so avoid confidential work.
Muse Spark 1.3 is Meta's multimodal reasoning model for complex agentic tasks. It accepts text, images, video, audio, and PDF documents, returns text, and has a 1M-token context window. Audio understanding is not fully supported, so use Muse Spark 1.2 for audio.
Muse Spark Open is Meta's open-weight, coding-optimized model for agentic coding. It is a sparse mixture-of-experts with 816B total and about 42B active parameters and a 1M-token context window. It reads text, images, video, audio and files and returns text.
Muse Wonder is Meta's frontier reasoning model for complex personal and professional agentic tasks, in preview. It accepts text, images, video, audio and PDF documents, returns text, and has a 1M-token context window.
Microsoft-Decision-1 is a small 9B Microsoft model built for fast decisions on yes/no, multiple-choice, and rating questions. It returns a calibrated probability for each fixed answer option, not generated chat text. It has a 32,768-token context window.
MiniMax M2.5 is MiniMax's large language model for real-world productivity. It extends M2.1 coding expertise into office work such as generating Word, Excel and PowerPoint files. It scores 80.2% on SWE-Bench Verified and 76.3% on BrowseComp.
MiniMax M2.7 is MiniMax's large language model for complex software engineering, agentic tool use and office productivity workflows. It supports complex agent harnesses, dynamic tool search and Agent Teams. It takes text and supports function calling.
MiniMax M2 is a large language model for complex software engineering, agentic tool use, and office productivity workflows. Supports complex agent harnesses, dynamic tool search, Agent Teams, and high-fidelity coding and document-editing tasks.
MiniMax M3 is MiniMax's multimodal language model for coding and agentic work. It accepts text, image and video input and returns text. It supports reasoning, function calling, structured output and multilingual use.
MiniMax M3.1 Flash is MiniMax's open-weight multimodal model for complex agentic tasks, coding and cybersecurity. It accepts text, image and video input and has a 1M-token context window. It always thinks, with adjustable effort, and its weights are not yet released.
Mistral Large 3 is Mistral AI's open-weight multimodal Mixture-of-Experts model with 675B total and 41B active parameters. It takes text and image input, supports tool use and JSON output, and has a 256K-token context. License is Apache 2.0.
Mistral Large 4 is Mistral AI's open-weight Mixture-of-Experts model with about 1 trillion parameters and a 1M-token context. It takes text and image input and supports tool use, structured output and Low, Medium or High reasoning effort.
Mistral Medium 3.5 128B is an open-weight 128B-parameter Mixture-of-Experts model balancing capability and efficiency. It is strong at reasoning, coding and multilingual tasks, and supports function calling with text input and output.
Mistral Small 3.1 24B Instruct builds on Mistral Small 3 (2501) and adds vision understanding. It handles long context up to 128k tokens, has 24 billion parameters, takes text and image input and supports function calling.
Mistral Small 3.2 24B Instruct is an open-weight 24B instruction model under Apache 2.0, with a 131k context window. It improves instruction following and function calling over Mistral Small 3.1 and takes text and image input.
Mistral Small 4 is a hybrid instruct-and-reasoning mixture-of-experts model with a 256k context window. It takes text and image input under Apache 2.0. It folds Mistral's separate Instruct, reasoning and coding lines into one model.
Moondream 3 Preview is Moondream's vision-language model with a mixture-of-experts architecture, 9B total and 2B active parameters. It answers questions about images, captions them and reads text, with a 32K-token context. Moondream 3.1 is its successor.
Moondream 3.1 9B A2B is Moondream's mixture-of-experts vision-language model with 9B total and 2B active parameters. It handles object detection, pointing, OCR and structured output, with a 32,768-token context window and up to 28,672 output tokens.
Kimi K2.5 is Moonshot AI's open-weight mixture-of-experts model with native vision, tool calling and structured outputs. It has 1T total and 32B active parameters, a 256K context window and a Modified MIT license.
Kimi K2.6 is Moonshot AI's open-weight large mixture-of-experts model with agentic tool use, vision input and long-context memory. It supports reasoning and streaming, and takes text and image input.
Kimi K2.7 Code is Moonshot AI's open-weight coding-focused model for end-to-end programming tasks over long contexts. It is a native multimodal mixture-of-experts model with 32B active of roughly 1T total parameters and always-on thinking.
Kimi K3.1 Pro is the first model in Moonshot AI's open-weight K3.1 series, a post-trained version of Kimi K3. It has a 1M-token context window and native visual understanding, and improves on K3 in coding, 3D and CAD modeling and financial research. Thinking is always on.
Kimi K3 is Moonshot AI's open-weight multimodal reasoning model with native visual understanding. It accepts text, image and video input for long-horizon software engineering, knowledge work and CAD. Thinking is always on.
Motif 3 is Motif Technologies' open-weight Mixture-of-Experts model, with 314B total and 13.2B active parameters. It is a general-purpose English and Korean model with a 262K context window, and it is strongest on agentic tool use and terminal tasks.
Drex Max v1.1 is Nace AI's largest multimodal decision model, the successor to Drex Max and Drex v2.0. It takes application state, up to 8 images and typed questions and returns calibrated probabilities for every answer option in one pass. It generates no text.
Drex v1.5 is a small decision model from Nace AI. It takes application state and typed questions and returns a calibrated probability for each fixed answer option, not generated chat text. It has a 131,072-token context window.
Drex v2.0 is Nace AI's 35B mixture-of-experts decision model with 3B active parameters, the larger successor to Drex v1.5. It returns a calibrated probability for every answer option in a single forward pass. It suits routing, classification and policy checks.
Nex-N2.5-Mini is Nex AGI's smaller agentic coding model, a MoE with 35B total and 3B active parameters. It is built for high-speed instruction following, real-time tool execution, and visually assisted computer use.
Nex-N2.5-Pro is Nex AGI's larger agentic coding model, a MoE with 397B total and 17B active parameters. It is built to turn goals into working, verified outcomes with a visual feedback loop and takes text and image input.
Nex-N2.6 is Nex-AGI's open-weight multimodal agentic model, built on a GLM Flash base model with about 321B parameters. It handles computer use, coding and long-horizon tool tasks, with text and image input and a 262K context window.
Hermes 3 405B Instruct is NousResearch's flagship open-weight model for advanced reasoning and coding, at 405B parameters. It supports function calling and streaming, with text input and output.
Nemotron-3 Nano 30B is a compact 30B-parameter text model from NVIDIA's Nemotron-3 family. It is tuned for fast, efficient text generation and understanding, with chat and streaming.
Nemotron-3 Nano Omni 30B-A3B (Reasoning) is NVIDIA's multimodal mixture-of-experts model for video, audio, image and text understanding. It targets enterprise Q&A, summarization, transcription and OCR, and supports reasoning and function calling.
Nemotron-3 Super 120B is NVIDIA's 120B-parameter reasoning model from the Nemotron-3 family. It supports tool calling and coding, and takes text input and returns text.
Nemotron 3 Ultra 550B is NVIDIA's 550B-parameter frontier model with 55B active, on a LatentMoE hybrid of Mamba-2, MoE and Attention. It has configurable reasoning through enable_thinking. It suits agentic workflows, long-context analysis and tool use.
Nemotron 3.5 Content Safety is NVIDIA's open-weight 4B-parameter multimodal guardrail model, fine-tuned from Google Gemma-3-4B. It moderates inputs to and responses from LLMs and VLMs. It returns a safety classification, category labels and optional reasoning.
Nemotron 3.5 Lightning 30B A3B is NVIDIA's sparse mixture-of-experts chat model, optimized for low-latency agentic workloads. It supports function calling and coding. It allows up to 131,072 output tokens.
Nemotron 3.5 Super VL 120B A12B is NVIDIA's vision-language model for reasoning, coding and tool use. It is a hybrid Mamba-2 and Transformer mixture-of-experts with 120B total and 12B active parameters. It takes text, image and video input.
GPT-4.1 Mini is OpenAI's smaller, faster version of GPT-4.1 with strong instruction following and tool calling. It has a 1,047,576-token input context and 32,768 output tokens. It takes text and image input and supports streaming, function calling and structured outputs.
GPT-4.1 Nano is OpenAI's fastest, most cost-efficient GPT-4.1 model, built for low-latency instruction following and tool calling at high volume. It has a 1,047,576-token input context and 32,768 output tokens, with text and image input.
GPT-4.1 is OpenAI's smartest non-reasoning model, built for instruction following and tool calling. It has a 1,047,576-token input context and 32,768 output tokens, and takes text and image input. It supports streaming, function calling and structured outputs.
GPT-4o Mini is OpenAI's fast, small multimodal model for focused tasks. It takes text and image input and returns text, with a 128,000-token input context and 16,384 output tokens. It supports streaming, function calling and structured outputs.
GPT-4o is OpenAI's versatile flagship multimodal model, with the o standing for omni. It takes text and image input and returns text, with a 128,000-token input context and 16,384 output tokens. It supports streaming, function calling and structured outputs.
GPT-5 Mini is OpenAI's reasoning model for low-latency, high-volume workloads. It allows 128,000 output tokens and takes text and image input. OpenAI recommends GPT-5.4 Mini for most new low-latency workloads.
GPT-5 Nano is OpenAI's fastest GPT-5 model, built for summarization, classification and other simple high-volume tasks. It allows 128,000 output tokens. OpenAI recommends GPT-5.4 Nano for new workloads.
GPT-5 Pro is OpenAI's version of GPT-5 that uses more compute to think harder on difficult tasks. It accepts text and image input, with a 400,000-token context window and up to 272,000 output tokens. It supports tool use and structured outputs.
GPT-5.1 Codex Max is OpenAI's maximum-effort variant of GPT-5.1 Codex for long-horizon agentic coding in the Codex environment. It has extended context and reasoning support, accepts text and image input, and supports streaming and function calling.
GPT-5.1 Codex Mini is OpenAI's smaller version of GPT-5.1 Codex for agentic coding where efficiency matters more than peak capability. It has a 400,000-token input context and 128,000 output tokens, with text and image input. It supports function calling and reasoning tokens.
GPT-5.1 Codex is OpenAI's version of GPT-5.1 for agentic coding in the Codex environment. It has a 400,000-token input context and 128,000 output tokens, with text and image input. It supports reasoning tokens, streaming, function calling and structured outputs.
GPT-5.1 is OpenAI's flagship model for coding and agentic tasks, with configurable reasoning effort from none to high. It allows 128,000 output tokens and takes text and image input. It supports streaming, function calling and structured outputs.
GPT-5.2 Codex is OpenAI's Codex-environment variant of GPT-5.2, tuned for agentic coding workloads. It accepts text and image input and supports reasoning, streaming, function calling and structured outputs.
GPT-5.2 Pro is OpenAI's version of GPT-5.2 that uses more compute to think harder on difficult tasks. It accepts text and image input, with a 400,000-token context window and up to 128,000 output tokens. It supports tool use and structured outputs.
GPT-5.2 is OpenAI's previous-generation frontier model for professional work, with configurable reasoning effort. It allows 128,000 output tokens and takes text and image input. OpenAI recommends GPT-5.4 for new workloads.
GPT-5.3 Codex is OpenAI's Codex-environment variant of GPT-5.3 for agentic coding. It accepts text and image input and supports reasoning, streaming, function calling and structured outputs.
GPT-5.4 Mini is OpenAI's mini model for coding, computer use and subagents, built for high-volume workloads. It allows 128,000 output tokens and takes text and image input. It supports reasoning tokens, function calling and structured outputs.
GPT-5.4 Nano is OpenAI's GPT-5.4-class model for simple, high-volume tasks such as classification, extraction and ranking. It allows 128,000 output tokens and takes text and image input. It supports function calling and structured outputs.
GPT-5.4 Pro is OpenAI's high-capacity model in the GPT-5 family, with advanced reasoning, broad knowledge and coding proficiency. It accepts text and image input and returns text, with tool-use capabilities.
GPT-5.4 is OpenAI's frontier model for complex professional workloads, with configurable reasoning effort from none to xhigh. It allows 128,000 output tokens and takes text and image input. It supports computer use, MCP and tool search.
GPT-5.5 Pro is OpenAI's Pro version of GPT-5.5, built for agentic workflows with built-in tools, improved reasoning and stateful context management. It accepts text and image input and supports coding, streaming and function calling.
GPT-5.5 is OpenAI's flagship model for coding, reasoning and multimodal work, with configurable reasoning effort from none to xhigh. It accepts text and image input. It supports streaming and function calling.
GPT-5.6 Luna is OpenAI's reasoning-focused model in the GPT-5.6 family, built for cost-sensitive workloads. It accepts text and image input. It supports streaming and function calling.
GPT-5.6 Sol is OpenAI's reasoning-focused model in the GPT-5.6 family, with configurable reasoning effort and image understanding. It accepts text, image and file input. It supports streaming and function calling.
GPT-5.6 Terra Pro is OpenAI's GPT-5.6 Terra served with reasoning mode set to pro for higher-quality answers on complex tasks. Pro mode spends far more reasoning tokens than standard GPT-5.6 Terra. It has a 1.05M-token context window and image input.
GPT-5.6 Terra is OpenAI's reasoning-focused model in the GPT-5.6 family that balances intelligence and cost. It accepts text and image input. It supports streaming and function calling.
GPT-5 is OpenAI's previous flagship reasoning model, built for coding and agentic tasks with configurable reasoning effort. It allows 128,000 output tokens and takes text and image input. OpenAI recommends GPT-5.1 for new workloads.
GPT-6 Astra Pro is OpenAI's GPT-6 Astra served with reasoning mode set to pro for higher-quality answers on complex tasks. Pro mode spends far more reasoning tokens than standard GPT-6 Astra. It accepts text, image and file input and supports function calling.
GPT-6 Astra is OpenAI's reasoning and chat model. It accepts text and image input and supports streaming and function calling.
GPT-6 Luna Decisions is GPT-6 Luna served through OpenAI's Decisions API, returning scores and choices instead of chat text. Each typed question gets a probability, a fixed-set choice or a rubric score. It has a 1,050,000-token context window and takes several questions at once.
GPT-6 Luna Pro is OpenAI's GPT-6 Luna served with reasoning mode set to pro for higher-quality answers on complex tasks. Pro mode spends far more reasoning tokens than standard GPT-6 Luna. It accepts text, image and file input and supports function calling.
GPT-6 Luna is OpenAI's efficient GPT-6 tier for focused, high-throughput tasks. It has configurable reasoning, tool use, image input. It supports streaming and function calling.
GPT-6 Sol Pro is OpenAI's GPT-6 Sol served with reasoning mode set to pro for higher-quality answers on complex tasks. Pro mode spends far more reasoning tokens than standard GPT-6 Sol. It accepts text, image and file input and supports function calling.
GPT-6 Sol is OpenAI's model for complex coding and agentic workflows. It has configurable reasoning, tool use, image input. It is the high-end GPT-6 tier below GPT-6 Astra.
GPT-6.1 Sol Pro is OpenAI's GPT-6.1 Sol served with reasoning mode set to pro for higher-quality answers on complex tasks. Pro mode spends far more reasoning tokens than standard GPT-6.1 Sol. It accepts text, image and file input and supports function calling.
GPT-6.1 Sol is OpenAI's model for complex coding and agentic work, with performance close to GPT-6 Astra. It has configurable reasoning, tool use, and image input.
GPT-6.1 Terra Pro is OpenAI's GPT-6.1 Terra served with reasoning mode set to pro for higher-quality answers on complex tasks. Pro mode spends far more reasoning tokens than standard GPT-6.1 Terra. It has a 1.05M-token context window and image input.
GPT-6.1 Terra is OpenAI's fast, efficient model for visual reasoning, precise computer use and clear, natural writing. It has configurable reasoning, tool use, image input and a 1.05M-token context window.
GPT-OSS-120b is OpenAI's open-weight mixture-of-experts reasoning model under the Apache-2.0 license. It has 116.8B total and 5.1B active parameters and a 131,072-token context window. It supports function calling and runs on one 80 GB GPU with MXFP4 weights.
GPT-OSS 20B is OpenAI's open-weight reasoning model for agentic tasks and low-latency edge use, under Apache-2.0. It is a mixture-of-experts model with 20.9B total and 3.6B active parameters and a 131,072-token context. It runs within 16 GB of memory with MXFP4 weights.
GPT-OSS Safeguard 20B is OpenAI's open-weight safety reasoning model built on gpt-oss-20b (Apache 2.0). It is a 21B-parameter MoE tuned for content classification, input and output filtering, and trust and safety judgments with reasoning traces.
o3-pro is OpenAI's version of o3 that uses more compute to think longer on hard problems. It accepts text and image input, with a 200,000-token context window and up to 100,000 output tokens. It supports tool use and structured outputs.
o3 is OpenAI's reasoning model for math, science, coding and visual analysis. It thinks before it answers and accepts text and image input. It supports tool use and structured outputs, with a 200,000-token context window and up to 100,000 output tokens.
text-embedding-3-large is OpenAI's text embedding model for English and non-English text. It returns 3,072-dimension vectors by default, which the dimensions parameter can shorten, and accepts up to 8,191 input tokens.
text-embedding-3-small is OpenAI's small text embedding model. It returns 1,536-dimension vectors by default, which the dimensions parameter can shorten, and accepts up to 8,191 input tokens.
Free Models Router is OpenRouter's router that selects free models at random from those available. It filters for features like image understanding, tool calling, and structured outputs. It accepts text and image input and has a 200K-token context window.
OpenRouter: Fusion is a multi-model deliberation router. A panel of expert models analyzes the prompt in parallel with web search and web fetch, then a judge model synthesizes a structured analysis and writes the final answer. It suits research and expert critique.
Parallel Decisions Fast is a web-grounded Decisions model. It takes text or JSON state and up to 100 typed questions, then returns structured answers with per-question refusals and citations after live web research. It does not generate chat text.
Particle-1 is Particle AI's token-efficient agentic coding model, post-trained on DeepSeek V4.1 Flash. It targets terminal workflows and codebase changes, with a 1M-token context window, image input, tool calling and reasoning.
Laguna XS 2.1 is Poolside's latest coding agent model in the 33B-A3B category, a step forward from Laguna XS.2. It combines tool calling and reasoning, with a 256K context window, up to 32K output tokens and FP8 quantization, under the OpenMDW-1.1 License.
Ternary Bonsai 2 27B is PrismML's 27B reasoning model, derived from Qwen3.8-27B and compressed to ternary weights. It supports coding, math, tool calling and image input, with a 262K-token context window.
Ternary Bonsai 2 9B is PrismML's compact open-weight vision-language model, derived from Qwen3.5-9B with ternary weights. It handles conversation, reasoning and tool use, takes image input and has a 262K-token context window.
Qwen Plus is Alibaba Cloud's balanced general-purpose Qwen text model on Model Studio. It covers text understanding, generation and reasoning at a moderate price, with tiered pricing by input length.
Qwen2.5-72B-Instruct is Alibaba Qwen's open-weight 72.7B-parameter general chat model. It is stronger at coding and maths than Qwen2, supports 29+ languages and handles function calling.
Qwen 2.5 Coder 32B is Alibaba's open-weight specialized coding model with 32B parameters. It targets code generation, completion and debugging across 92 programming languages, with function calling and streaming.
Qwen3-14B is an open-weight 14.8B-parameter dense model from Alibaba Qwen that balances quality and latency. It supports reasoning, function calling, 100+ languages and dialects, instruction following and translation.
Qwen3 235B A22B (Instruct 2507) is Qwen's open-weight, non-thinking Mixture-of-Experts model with 235B total and 22B active parameters. It is licensed Apache-2.0 and covers 119 languages and dialects, coding and tool use.
Qwen3 Coder Flash is the fast coding model of Alibaba's Qwen3-Coder family on Model Studio. It trades some depth for speed in code completion and agentic coding tools.
Qwen3 Embedding 4B is the open-weight 4B size in the Qwen3 embedding and reranking family. It embeds text over a 32,768-token window at 2,560 dimensions, and supports 100+ languages and Matryoshka truncation from 32 to 2,560 dimensions.
Qwen3.5-27B is Alibaba's largest open-weight dense Qwen3.5 model. It has a thinking mode, tool calling, multi-token prediction and 201 languages, with text, image and video input.
Qwen 3.5 397B A17B is Alibaba's open-weight mixture-of-experts model with 397B total and 17B active parameters. It supports reasoning and function calling, and takes text, image and video input.
Qwen3.5-9B is an open-weight model from Alibaba's Qwen3.5 series with a hybrid Gated Delta Networks and sparse MoE architecture. It has a thinking mode, tool calling and 201 languages.
Qwen3.6 Flash is the fast, low-cost native vision-language model of Alibaba's Qwen3.6 series on Model Studio. It improves on Qwen3.5 Flash in agentic coding, math and code reasoning, spatial understanding and object localization.
Qwen3.6 Plus is Alibaba's plus-tier multimodal model of the Qwen3.6 generation. It takes text and image input, has a 1 million token context window, and supports reasoning, coding and tool use.
Qwen3.7 Flash is Alibaba's fast vision-language reasoning model for multimodal agents. It suits visual coding, search and computer interaction, with strengths in object recognition, spatial understanding and real-world visual perception.
Qwen3.7 Max is Alibaba's flagship max-tier model of the Qwen3.7 generation. It takes text and image input, has a 1 million token context window, and supports reasoning, coding and tool use.
Qwen3.7 Plus is Alibaba's advanced multimodal model with text and image input. It has a 1 million token context window and strong reasoning capabilities.
Qwen3.8 2.4T A95B is Qwen's sparse mixture-of-experts model, the open-weight variant of Qwen3.8 Max. It has 95 billion active of 2.4 trillion total parameters.
Qwen3.8-27B is Alibaba's open-weight dense Qwen3.8 model for agentic coding and complex reasoning, with a native thinking mode. It targets multi-step agent work and large coding tasks.
Qwen3.8 Flash is Alibaba Qwen's native vision-language model for coding, office work, long-context reasoning and agents. It has a 1M context window, takes text, image and video input, and supports thinking and function calling.
Qwen3.8 Max Prime is a higher-throughput variant of Qwen3.8 Max from Alibaba's Qwen team. It takes text, image and video input and supports reasoning and function calling.
Qwen3.8 Max is Alibaba's flagship 2.4-trillion parameter MoE model of the Qwen3.8 generation. It takes text, image and video input, has a 1 million token context window, and supports reasoning, coding and tool use.
Qwen3.8 Omni Flash is Alibaba Cloud Qwen's native multimodal model for text, image, audio and video input at 1M context. It is built for agentic programming, knowledge work, GUI operation and audio and video workflows.
Flint is Recursive's open-weight mixture-of-experts model for coding agents, with 39.8B total and 2.5B active parameters. It supports reasoning and multi-turn tool use with a 262,144-token context. It is built for fixing bugs and writing tests.
Beam 1.0 is Reflection's open-weights mixture-of-experts model with 501B total and 23B active parameters. It is built for token-efficient reasoning, agentic tool use, and software engineering. Weights are Apache-2.0.
Span-01 Lite is Respan's behavior-scoring classifier, the lighter tier of Span-01, a 4B model trained with RLAIF. It takes a conversation span and plain-language behavior definitions and returns a present, absent or not_observable probability per behavior in one forward pass.
Span-01 is Respan's behavior-scoring classifier, a 4B model trained with RLAIF on classification reasoning rather than fixed taxonomies. It returns a present, absent or not_observable probability per behavior in one forward pass. Span-01 Lite is the lighter tier.
Fugu Max is the largest tier of Sakana AI's Fugu line. It has a 1M-token context and takes text, images, and files as input and returns text. It is a distinct tier alongside Fugu and Fugu Ultra.
Fugu Ultra v2 is the higher-performance model in Sakana AI's Fugu family, a learned multi-agent orchestration system. It takes text, images, and files as input and returns text, with reasoning and function calling.
Fugu Ultra is Sakana AI's flagship orchestration model. It coordinates a deep pool of expert agents and frontier models behind one OpenAI-compatible API, then synthesizes the result. It supports Chat Completions, Responses, and web_search.
Sakana Namazu is Sakana AI's Japanese-specialized LLM, built on the open model Kimi K2.6 and tuned for Japanese and Japanese business contexts. It combines web search and code execution to carry complex tasks through to completion.
L3.1 Euryale 70B is Sao10k's text-generation model for creative roleplay. It is the successor to Euryale L3 70B v2.1 and takes text input with streamed text output.
Classify-1 is ScaleDown's structured decision model that scores text or a document against choice, yes/no and ordered-scale questions. It returns calibrated probabilities instead of generated prose, suited to routing, moderation, intent detection and domain tagging.
Extract is ScaleDown's description-driven entity extraction model. Callers describe the fields they want in plain English, and it returns each matching span with a confidence score, character offsets and context. Built-in OCR handles scanned images and PDFs.
Summarize is ScaleDown's abstractive summarization model that condenses long text and documents into a faithful rewrite. It suits document review, transcript digests and RAG pre-processing, with format, focus, language and length steered through the system prompt.
Step 5 Preview is StepFun's open-weight flagship base model for real-world tasks, with a focus on programming and professional expertise. It is multimodal, taking text, image and video in and text out, with a 1M-token context and adjustable reasoning effort.
Step 3.5 Flash is StepFun's open-weight sparse Mixture-of-Experts model with 196.81B total parameters and about 11B active per token. It uses a 45-layer transformer with 288 routed experts, Top-8 selection, and 256K context for reasoning and agentic tasks.
Step 3.7 Flash is StepFun's open-weight 1T parameter multimodal Mixture-of-Experts model for long-horizon coding, agentic tool use, and image and video understanding. It takes text and image input and supports structured output.
Hy-MT2 1.8B is the smallest model in Tencent's Hy-MT2 translation family. It translates among 33 languages and follows translation instructions, with an 8,192-token window. It is open-weight under Apache-2.0.
Hy-MT2 30B A3B is Tencent's open-weight translation model, a 30B-parameter MoE with 3B active parameters. It is tuned for machine translation with an 8,192-token window. It is the largest of the published Hy-MT2 sizes.
Hy4 Preview is Tencent Hunyuan's open-weight 770B mixture-of-experts model with 49B active parameters, built for agents and coding. It targets office automation and complex tool use.
Hy4 is Tencent Hunyuan's 1M-token-context model for agents, coding and long-horizon work, the official release after Hy4 Preview. It accepts text and image input, supports tool calling and structured outputs, and has optional reasoning on by default.
Inkling 2 Small is Thinking Machines Lab's second-generation multimodal model. It is a 594B-parameter mixture-of-experts transformer with 17B active, a 1M-token context, reasoning effort, and tool use. Weights are closed under the Tinker Terms of Use.
Inkling Small is an open-weight multimodal mixture-of-experts model from Thinking Machines Lab. It has 12B active parameters out of 276B total. It takes text, image, and audio input and returns text.
Inkling is Thinking Machines Lab's first open-weights foundation model. It is a 975B-parameter multimodal mixture-of-experts transformer with 41B active. It takes text, image, and audio input, with switchable reasoning effort and tool use.
Jev is TypeSafe's decision model that takes application state and typed questions and returns structured answers with probabilities. It returns decisions rather than generated chat text.
Pareto 26.10 is Unbiased's composite model that blends several language models into one answer. It takes text and image input and returns text over a 1,048,576-token context window, with 131,072 max completion tokens. It is aimed at research, coding and agentic workflows.
Pareto 26.9 is the September 2026 release of Unbiased's blended model that combines frontier and open models into one answer. It takes text and image input with a 262,144-token context. It targets research, coding and agentic workflows and was superseded by Pareto 26.10.
Solar Decide Flash is Upstage's low-latency variant of Solar Decide, a decision model built on Solar Mini 4. It takes application state and typed noul, choice or score questions and returns structured answers, not chat text. Context is 524,288 tokens.
Solar Decide is Upstage's decision model built on Solar Mini 4. It takes application state and typed noul, choice or score questions and returns structured answers, not generated chat text. It has a 524,288-token context window.
Solar Mini 4 is Upstage's small, fast tier in the Solar 4 generation. It is a text-only model with a 524,288-token context window. It supports function calling and reasoning, and is a separate model from Solar Pro 4.
Solar Pro 4 is Upstage's cost-efficient large language model with a 524K context window. It is built for long-horizon tasks and agentic workflows, and supports function calling and reasoning with text input and output.
xAI's Grok 4.20 multi-agent snapshot (March 2026), optimized for agentic workflows, tool use, and coordinated multi-step task planning and execution.
xAI's Grok 4.20 non-reasoning snapshot (March 2026), optimized for fast standard conversational and analytical tasks without extended chain-of-thought reasoning.
xAI's Grok 4.20 reasoning snapshot (March 2026) featuring extended step-by-step thinking for complex problem-solving, mathematics, and scientific reasoning tasks.
xAI's Grok 4.3 is an iterative update from the Grok 4 series offering improved instruction-following, general reasoning, and broad conversational capabilities.
Grok 4.5 is SpaceXAI's model with frontier performance on coding, knowledge work, and STEM. It takes text and image input, returns text, and supports function calling and streaming.
Grok 4.6 is SpaceXAI's model with frontier performance on coding, knowledge work, and STEM. It builds on Grok 4.5 for long-running agents and visual work, with text and image input and function calling.
Grok 4.7 is SpaceXAI's frontier model for coding, agentic tasks, and knowledge work. It succeeds Grok 4.6 with mandatory adjustable reasoning at low, medium, high, or xhigh effort, defaulting to high. It takes text and image input.
Grok Build 0.1 is xAI's coding and automation model. It takes text and image input and supports function calling, structured outputs, and reasoning.
MiMo V2.5 Pro is Xiaomi's open-weight flagship model for agentic work, complex software engineering, and long-horizon tasks. It accepts text, image and video input. It completes professional tasks involving thousands of tool calls.
MiMo V2.5 is Xiaomi's open-weight native omnimodal model with Pro-level agentic performance. It accepts text, image and video input, improves on MiMo V2 Omni in multimodal perception, and suits agent frameworks.
MiMo V2.6 Flash is Xiaomi's full-modality reasoning model for high-frequency invocations and large-scale professional workloads. It keeps the 128K output limit of MiMo V2.6 Pro, with text, image and video input.
MiMo-V2.6-Pro UltraSpeed is the fast edition of Xiaomi's flagship MiMo-V2.6-Pro, built from the same 1T checkpoint with roughly 10x faster output. It accepts text, image, video and audio input and has a 1M-token context window.
MiMo V2.6 Pro is Xiaomi's trillion-parameter, natively omni-modal flagship reasoning model. It targets complex projects, long-horizon agentic tasks, coding, research, and professional knowledge work, with up to 128K output tokens.
Xing4.0 29B A4B is an open-weight mixture-of-experts reasoning model from China Telecom's TeleAI. It activates 4B of 29B parameters per token and supports a 256K context. It targets agentic coding, tool calling, and long-context engineering tasks.
GLM-4.5 is Z-AI's open-weight foundational model for agent-oriented applications. It is a mixture-of-experts model with 355B total and 32B active parameters, trained on 15 trillion tokens. It has a 128K context window, hybrid thinking modes, and strong tool use and coding.
GLM-4.6 is Z-AI's open-weight mixture-of-experts model with 355B total and 32B active parameters, under the MIT license. It has a 200K context window, up from 128K, and a 128K max output. It improves coding, tool use, and search-based agents.
GLM-4.6V is the vision open-weight variant of Z-AI's GLM-4.6 model. It takes text and image input. It suits document intelligence, code-screenshot interpretation, and multimodal reasoning.
GLM-4.7 Flash is an open-weight fast multilingual text generation model from Z-AI with a 131,072 token context window. It is optimized for dialogue, instruction following, and multi-turn tool calling across 100+ languages.
GLM-4.7 is Z-AI's open-weight multilingual text generation model with a 131,072 token context window. It supports chat, instruction following, multi-turn tool calling, and reasoning across 100+ languages.
GLM-5-Turbo is Z.ai's fast GLM-5 variant, tuned for agent workloads such as OpenClaw. It handles tool calling, instruction following, and scheduled and long-running tasks. It has a 200K-token context window, up to 128K output tokens, and thinking mode.
GLM-5.1 is Z-AI's open-weight flagship foundation model for long-horizon autonomous tasks. It has a 128K output window and scores 58.4 on SWE-Bench Pro. It supports thinking mode, function calling, and MCP integration.
GLM-5.2 is Z-AI's open-weight flagship reasoning model with a 128K output window. It supports controllable thinking effort, function calling, and MCP integration. It targets long-horizon autonomous coding and agentic tasks.
GLM-5.3-FlashX-Legal is Z.ai's model for professional legal workflows, built on GLM-5.3-FlashX. It covers legal research, contract review, and document drafting. It takes text, image, video, and file input, returns text, and has a 1M-token context window.
GLM-5.3-FlashX is Z.AI's high-speed inference variant for coding agents, real-time interactions, and long-running agentic workflows. It takes text, image, and video input and returns text.
GLM 5.3 Prime is the top reasoning tier of Z.ai's GLM 5.3 line, above the Flash and FlashX cuts. It has a 1M-token context window and supports function calling and streaming with text in and text out.
GLM-5.3 is Z.AI's open-weight reasoning model for coding and agentic workflows. It targets complex software engineering, long-running agents, and vulnerability analysis. It uses the same base model as GLM-5.2 with scaled post-training for stronger coding and token efficiency.
GLM-5 is Z-AI's open-weight general-purpose foundation model for reasoning, coding, and agentic workloads. It has a 128K context window and 32K output. It supports thinking mode, function calling, and structured outputs.