API Status Page

The other really good open source model from China. Alibaba's Qwen is very similar to Llama. Good on its own, but strongest when distilled by other data sets or models.

Qwen3-235B-A22B-Thinking-2507 is a high-performance, open-weight Mixture-of-Experts (MoE) language model optimized for complex reasoning tasks. It activates 22B of its 235B parameters per forward pass and natively supports up to 262,144 tokens of context. This "thinking-only" variant enhances structured logical reasoning, mathematics, science, and long-form generation, showing strong benchmark performance across AIME, SuperGPQA, LiveCodeBench, and MMLU-Redux. It enforces a special reasoning mode (</think>) and is designed for high-token outputs (up to 81,920 tokens) in challenging domains. The model is instruction-tuned and excels at step-by-step reasoning, tool use, agentic workflows, and multilingual tasks. This release represents the most capable open source variant in the Qwen 3 series, surpassing many closed models in structured reasoning use cases.

Qwen3-235B-A22B-Instruct-2507 is a multilingual, instruction-tuned mixture-of-experts language model based on the Qwen3-235B architecture, with 22B active parameters per forward pass. It is optimized for general-purpose text generation, including instruction following, logical reasoning, math, code, and tool usage. The model supports a native 262K context length and does not implement "thinking mode" (<think> blocks). Compared to its base variant, this version delivers significant gains in knowledge coverage, long-context reasoning, coding benchmarks, and alignment with open-ended tasks. It is particularly strong on multilingual understanding, math reasoning (e.g., AIME, HMMT), and alignment evaluations like Arena-Hard and WritingBench.

Qwen 3-Coder is a cutting-edge model from Alibaba that rivals o3, Deepseek, and even Kimi K2. It's the current SOTA in agentic tool use and coding, and excels at complex reasoning and programming tasks, including code generation, reasoning, and tool use. Qwen 3-Coder is designed to be a versatile tool for developers, researchers, and students, offering a powerful combination of reasoning and code generation capabilities.

DeepSeek V3, a 685B-parameter, mixture-of-experts model, is the latest iteration of the flagship chat model family from the DeepSeek team. It succeeds the DeepSeek V3 model and performs really well on a variety of tasks.

DeepSeek-V3.1 is a large hybrid reasoning model (671B parameters, 37B active). It extends the DeepSeek-V3 base with a two-phase long-context training process, reaching up to 128K tokens, and uses FP8 microscaling for efficient inference. The model improves tool use, code generation, achieving performance comparable to DeepSeek-R1 on difficult benchmarks while responding more quickly. It succeeds the DeepSeek V3-0324 model and performs well on a variety of tasks.

DeepSeek-V3.1 is a large hybrid reasoning model (671B parameters, 37B active). It extends the DeepSeek-V3 base with a two-phase long-context training process, reaching up to 128K tokens, and uses FP8 microscaling for efficient inference. The model improves tool use, code generation, and reasoning efficiency, achieving performance comparable to DeepSeek-R1 on difficult benchmarks while responding more quickly. It succeeds the DeepSeek V3-0324 model and performs well on a variety of tasks. This version has reasoning capabilities enabled, allowing it to provide responses with greater accuracy and nuanced context handling.

DeepSeek-V3.1 Terminus is an update to DeepSeek V3.1 that maintains the model's original capabilities while addressing issues reported by users, including language consistency and agent capabilities, further optimizing the model's performance in coding and search agents. It is a large hybrid reasoning model (671B parameters, 37B active) that supports both thinking and non-thinking modes. It extends the DeepSeek-V3 base with a two-phase long-context training process, reaching up to 128K tokens, and uses FP8 microscaling for efficient inference. The model improves tool use, code generation, and reasoning efficiency, achieving performance comparable to DeepSeek-R1 on difficult benchmarks while responding more quickly. It supports structured tool calling, code agents, and search agents, making it suitable for research, coding, and agentic workflows.

DeepSeek-V3.1 Terminus is an update to DeepSeek V3.1 that maintains the model's original capabilities while addressing issues reported by users, including language consistency and agent capabilities, further optimizing the model's performance in coding and search agents. It is a large hybrid reasoning model (671B parameters, 37B active) that supports both thinking and non-thinking modes. It extends the DeepSeek-V3 base with a two-phase long-context training process, reaching up to 128K tokens, and uses FP8 microscaling for efficient inference. The model improves tool use, code generation, and reasoning efficiency, achieving performance comparable to DeepSeek-R1 on difficult benchmarks while responding more quickly. It supports structured tool calling, code agents, and search agents, making it suitable for research, coding, and agentic workflows.

DeepSeek-V3.2-Exp is an experimental large language model released by DeepSeek as an intermediate step between V3.1 and future architectures. It introduces DeepSeek Sparse Attention (DSA), a fine-grained sparse attention mechanism designed to improve training and inference efficiency in long-context scenarios while maintaining output quality. The model was trained under conditions aligned with V3.1-Terminus to enable direct comparison. Benchmarking shows performance roughly on par with V3.1 across reasoning, coding, and agentic tool-use tasks, with minor tradeoffs and gains depending on the domain. This release focuses on validating architectural optimizations for extended context lengths rather than advancing raw task accuracy, making it primarily a research-oriented model for exploring efficient transformer designs.

DeepSeek-V3.2-Exp is an experimental large language model released by DeepSeek as an intermediate step between V3.1 and future architectures. It introduces DeepSeek Sparse Attention (DSA), a fine-grained sparse attention mechanism designed to improve training and inference efficiency in long-context scenarios while maintaining output quality. The model was trained under conditions aligned with V3.1-Terminus to enable direct comparison. Benchmarking shows performance roughly on par with V3.1 across reasoning, coding, and agentic tool-use tasks, with minor tradeoffs and gains depending on the domain. This release focuses on validating architectural optimizations for extended context lengths rather than advancing raw task accuracy, making it primarily a research-oriented model for exploring efficient transformer designs.

The open source reasoning model that shook the whole industry. Very smart. Shows all of its thinking. Not the fastest.

The open source reasoning model that shook the whole industry. Very smart. Shows all of its thinking. Not the fastest. New weights released on 5/28/2025.

Similar to the Llama distilled model, but distilled on Qwen 32B instead. Slightly better at code, slightly more likely to fall into thought loops.

Google's flagship model, known for speed and accuracy (and also web search!). Not quite as smart as Claude 3.5 Sonnet, but WAY faster and cheaper. Also has an insanely large context window (it can handle a lot of data).

Google's state of the art fast model, known for speed and accuracy (and also web search!). Not quite as smart as Claude Sonnet, but WAY faster and cheaper. Also has an insanely large context window (it can handle a lot of data).

Google's state of the art fast model, known for speed and accuracy, now with support for "thinking". These "thinking" capabilities enable it to provide responses with greater accuracy and nuanced context handling.

Gemini 2.5 Flash-Lite is a member of the Gemini 2.5 series of models, a suite of highly-capable, natively multimodal models. Gemini 2.5 Flash-Lite is Google’s most cost-efficient model, striking a balance between efficiency and quality.

Gemini 2.5 Flash-Lite is a member of the Gemini 2.5 series of models, a suite of highly-capable, natively multimodal models. Gemini 2.5 Flash-Lite is Google’s most cost-efficient model, striking a balance between efficiency and quality. This version has "thinking" capabilities that enable it to provide responses with greater accuracy and nuanced context handling.

Gemini 2.5 Flash Image Preview is a state of the art image generation model with contextual understanding. It is capable of image generation, edits, and multi-turn conversations.

Similar to 2.0 Flash, but even faster. Not as smart, but still good at most things.

Google's most advanced model, excelling at complex reasoning and problem-solving. Particularly strong at tackling difficult code challenges, mathematical proofs, and STEM problems. With its massive context window, it can deeply analyze large codebases, datasets and technical documents to provide comprehensive solutions.

Google's Imagen 4 is a powerful image generation model that creates high-quality, photorealistic images from text prompts. Built on advanced diffusion techniques and trained on diverse datasets. 2 images per prompt.

Google's Imagen 4 Ultra is a powerful image generation model that creates high-quality, photorealistic images from text prompts. Built on advanced diffusion techniques and trained on diverse datasets. 1 image per prompt.

Google's latest flagship model excels at advanced reasoning and problem-solving. It's especially strong with complex code challenges, mathematical proofs, and STEM topics. Thanks to its vast context window, it can deeply analyze large codebases, datasets, and technical documents to deliver thorough, high-quality solutions.

Nano Banana Pro is a state of the art image generation model with contextual understanding. It is capable of image generation, edits, and multi-turn conversations.

Llama 4 Scout 17B Instruct (16E) is a mixture-of-experts (MoE) language model developed by Meta, activating 17 billion parameters out of a total of 109B. It supports native multimodal input (text and image) and multilingual output (text and code) across 12 supported languages. Designed for assistant-style interaction and visual reasoning, Scout uses 16 experts per forward pass and features a context length of up to 10 million tokens, with a training corpus of ~40 trillion tokens. Built for high efficiency and local or commercial deployment, Llama 4 Scout incorporates early fusion for seamless modality integration. It is instruction-tuned for use in multilingual chat, captioning, and image understanding tasks. Released under the Llama 4 Community License, it was last trained on data up to August 2024 and launched publicly on April 5, 2025.

Llama 4 Maverick 17B Instruct (128E) is a high-capacity multimodal language model from Meta, built on a mixture-of-experts (MoE) architecture with 128 experts and 17 billion active parameters per forward pass (400B total). It supports multilingual text and image input, and produces multilingual text and code output across 12 supported languages. Optimized for vision-language tasks, Maverick is instruction-tuned for assistant-like behavior, image reasoning, and general-purpose multimodal interaction. Maverick features early fusion for native multimodality and a 1 million token context window. It was trained on a curated mixture of public, licensed, and Meta-platform data, covering ~22 trillion tokens, with a knowledge cutoff in August 2024. Released on April 5, 2025 under the Llama 4 Community License, Maverick is suited for research and commercial applications requiring advanced multimodal understanding and high model throughput.

MiniMax-M2 is a compact, high-efficiency large language model optimized for end-to-end coding and agentic workflows. With 10 billion activated parameters (230 billion total), it delivers near-frontier intelligence across general reasoning, tool use, and multi-step task execution while maintaining low latency and deployment efficiency. The model excels in code generation, multi-file editing, compile-run-fix loops, and test-validated repair, showing strong results on SWE-Bench Verified, Multi-SWE-Bench, and Terminal-Bench. It also performs competitively in agentic evaluations such as BrowseComp and GAIA, effectively handling long-horizon planning, retrieval, and recovery from execution errors.

Kimi K2 is a large-scale Mixture-of-Experts (MoE) language model developed by Moonshot AI, featuring 1 trillion total parameters with 32 billion active per forward pass. It is optimized for agentic capabilities, including advanced tool use, reasoning, and code synthesis. Kimi K2 excels across a broad range of benchmarks, particularly in coding (LiveCodeBench, SWE-bench), reasoning (ZebraLogic, GPQA), and tool-use (Tau2, AceBench) tasks. It supports long-context inference up to 128K tokens and is designed with a novel training stack that includes the MuonClip optimizer for stable large-scale MoE training.

Kimi K2 0905 is the September update of Kimi K2 0711. It is a large-scale Mixture-of-Experts (MoE) language model developed by Moonshot AI, featuring 1 trillion total parameters with 32 billion active per forward pass. It supports long-context inference up to 256k tokens, extended from the previous 128k. Kimi K2 is optimized for agentic capabilities, including advanced tool use and code synthesis. It excels across coding (LiveCodeBench, SWE-bench), and tool-use (Tau2, AceBench) benchmarks. The model is trained with a novel stack incorporating the MuonClip optimizer for stable large-scale MoE training.

Kimi K2 Thinking is an update to Kimi K2 that is trained for reasoning and agentic tasks. It is a large-scale Mixture-of-Experts (MoE) language model developed by Moonshot AI, featuring 1 trillion total parameters with 32 billion active per forward pass. It supports long-context inference up to 256k tokens. Kimi K2 is optimized for agentic capabilities, including advanced tool use, reasoning, and code synthesis. It rivals GPT-5 and Claude 4.5 Sonnet across several benchmarks including HLE, BrowseComp, SWE-Multilingual, and LiveCodeBench.

Kimi K2 Thinking is an update to Kimi K2 that is trained for reasoning and agentic tasks. It is a large-scale Mixture-of-Experts (MoE) language model developed by Moonshot AI, featuring 1 trillion total parameters with 32 billion active per forward pass. It supports long-context inference up to 256k tokens. Kimi K2 is optimized for agentic capabilities, including advanced tool use, reasoning, and code synthesis. It rivals GPT-5 and Claude 4.5 Sonnet across several benchmarks including HLE, BrowseComp, SWE-Multilingual, and LiveCodeBench.

A medium-sized open-weight model from OpenAI suitable for general-purpose tasks. 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 and deployability on consumer or single-GPU hardware. 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.

A large open-weight model from OpenAI. gpt-oss-120b is an open-weight, 117B-parameter Mixture-of-Experts (MoE) language model from OpenAI designed for high-reasoning, agentic, and general-purpose production use cases. It activates 5.1B parameters per forward pass and is optimized to run on a single H100 GPU with native MXFP4 quantization. The model supports configurable reasoning depth, full chain-of-thought access, and native tool use, including function calling, browsing, and structured output generation.

Like gpt-4o, but faster. This model sacrifices some of the original GPT-4o's precision for significantly reduced latency. It accepts both text and image inputs.

OpenAI's flagship non-reasoning model. Works with text and images. Relatively smart. Good at most things.

GPT-4.1 is a flagship large language model optimized for advanced instruction following, real-world software engineering, and long-context reasoning. It outperforms GPT-4o and GPT-4.5 across coding (54.6% SWE-bench Verified), instruction compliance (87.4% IFEval), and multimodal understanding benchmarks.

GPT-4.1 Mini is a mid-sized model delivering performance competitive with GPT-4o at substantially lower latency. It has a very large context window and scores 45.1% on hard instruction evals, 35.8% on MultiChallenge, and 84.1% on IFEval. Mini also shows strong coding ability (e.g., 31.6% on Aider's polyglot diff benchmark) and vision understanding.

For tasks that demand low latency, GPT‑4.1 nano is the fastest model in the GPT-4.1 series. It delivers exceptional performance at a small size with its 1 million token context window, and scores 80.1% on MMLU, 50.3% on GPQA, and 9.8% on Aider polyglot coding – even higher than GPT‑4o mini. It's ideal for tasks like classification or autocompletion.

OpenAI's latest flagship model. PhD-level intelligence at most things. This is the version that is the best to chat with.

OpenAI's latest flagship model. PhD-level intelligence at most things. This version has reasoning capabilities, and is not well suitied to general chat, but is great for complex reasoning tasks.

A lighter-weight GPT-5 variant optimized for speed while retaining strong reasoning and tool use.

An ultra-fast GPT-5 variant tuned for low-latency tasks with reasoning and tool use.

GPT-5.1 Instant delivers a major step forward in conversational fluency, recall, and following user instructions, with noticeably lower latency compared to prior models. Ideal for real-time chat and rapid exchanges, it demonstrates greatly improved contextual understanding, memory, and following complex instructions. GPT-5.1 Instant also benefits from upgraded image understanding, web browsing, and code/model reasoning abilities.

GPT-5.1 Thinking is the flagship GPT-5.1, excelling in deep recall, tool use, and complex multi-step reasoning—especially over longer conversations or documents. It leverages the expanded context window, improved factual accuracy, and new instruction-following capabilities of the 5.1 release. This model shines on tasks that demand in-depth problem solving, retrieval, code, and multi-modal reasoning.

A small, fast, super smart reasoning model. OpenAI clearly didn't want DeepSeek to be getting all the attention. Good at science, math, and coding, even if it's not as good at CSS.

A small, fast, even smarter reasoning model. o3-mini was great, this is even better. Good at science, math, and coding, even if it's not as good at CSS.

o3 is a well-rounded and powerful model across domains. It sets a new standard for math, science, coding, and visual reasoning tasks. It also excels at technical writing and instruction-following. Use it to think through multi-step problems that involve analysis across text, code, and images.

The o3 series of models are trained with reinforcement learning to think before they answer and perform complex reasoning. The o3-pro model uses more compute to think harder and provide consistently better answers.

OpenAI's latest and greatest image generation model, using lots of crazy tech like custom tools for text and reflections. Best image gen available today by a mile.

GLM-4.5 is an open-weight MoE model that competes with o3 and Claude 4 while being smaller and stronger than DeepSeek-R1 and Kimi K2. It excels at reasoning, coding, and agentic tasks and is trained using the Muon architecture, the same one to train Kimi K2.

GLM-4.5 is an open-weight MoE model that competes with o3 and Claude 4 while being smaller and stronger than DeepSeek-R1 and Kimi K2. It excels at reasoning, coding, and agentic tasks and is trained using the Muon architecture, the same one to train Kimi K2. This variant has reasoning mode enabled for step-by-step thinking.

GLM-4.5V is an open-weight MoE model that competes with o3 and Claude 4 while being smaller and stronger than DeepSeek-R1 and Kimi K2. It excels at reasoning, coding, and agentic tasks and is trained using the Muon architecture, the same one to train Kimi K2.

GLM-4.5V is an open-weight MoE model that competes with o3 and Claude 4 while being smaller and stronger than DeepSeek-R1 and Kimi K2. It excels at reasoning, coding, and agentic tasks and is trained using the Muon architecture, the same one to train Kimi K2. This variant has reasoning mode enabled for step-by-step thinking.

GLM-4.5-Air is the lightweight variant of GLM-4.5, an open-weight MoE model that competes with o3 and Claude 4 while being smaller and stronger than DeepSeek-R1 and Kimi K2. It excels at reasoning, coding, and agentic tasks and is trained using the Muon architecture, the same one to train Kimi K2.

GLM-4.5 Air is the lightweight variant of GLM-4.5, an open-weight MoE model that competes with o3 and Claude 4 while being smaller and stronger than DeepSeek-R1 and Kimi K2. It excels at reasoning, coding, and agentic tasks and is trained using the Muon architecture, the same one to train Kimi K2. This variant has reasoning mode enabled for step-by-step thinking.

Compared with GLM-4.5, this generation 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.

Compared with GLM-4.5, this generation 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.

xAI's flagship model that breaks records on lots of benchmarks (allegedly). Possesses deep domain knowledge in finance, healthcare, law, and science.

xAI's flagship model that excels at data extraction, coding, and text summarization. Possesses deep domain knowledge in finance, healthcare, law, and science.

A lightweight model that thinks before responding. Great for simple or logic-based tasks that do not require deep domain knowledge.

Grok 4 Fast is xAI's latest multimodal model with SOTA cost-efficiency and a 2M token context window.

Grok 4 Fast is xAI's latest multimodal model with SOTA cost-efficiency and a 2M token context window.

Grok 4.1 Fast is xAI's latest multimodal model with SOTA cost-efficiency and a 2M token context window.

Grok 4.1 Fast is xAI's latest multimodal model with SOTA cost-efficiency and a 2M token context window.