LLM Models · Closed Source
Browse the world's large language models. Compare parameters, benchmarks, VRAM and more.
Closed-source listings are models you are expected to call as a service. You get an API and a vendor. You do not get the weights. That is the right trade when you need a result without operating a server, when the data may leave your network, and when the vendor's current model wins on a prompt you are allowed to send.
Read the company line and the exact name. Families have several rows: standard, fast, batch, and Latest. Fast is often the same model at a different price and speed. Batch is an endpoint type. Latest is a pointer. Compare the variant you will call. Deciding after a comparison tells you not to mix those up and then stare at highlighted cells.
Price on the page is a snapshot. Cite the vendor's pricing page in any document other people will rely on. Context window is the vendor's claimed cap, not a promise that the middle of a million tokens is attended to equally. If your prompt is short, stop spending attention on the window and look at price.
This category is not "safer" and not "better." It is "someone else runs it." If policy forbids that, move to open source even if the scores there are lower. The score did not read your policy. How ranking works says the score is a sort key.
Confirm data retention on the vendor's site before you send production prompts. Our card does not negotiate their terms. The terms of this directory do not either.
Put the vendor's data-retention sentence in the same note as the model id. If you cannot find that sentence, you are not ready to send production text. The directory card will not grow that sentence for you overnight. If the vendor moves the policy, your note is stale and the pin should be reread before the next release.
GPT-5.6 Sol Pro is the same underlying model as [GPT-5.6 Sol](https://openrouter.ai/openai/gpt-5.6-sol), served with `reasoning.mode` set to `pro` for higher-quality responses on complex tasks. Learn more in OpenAI's docs: https://developers.openai.com/api/docs/guides/reasoning#reasoning-mode
GPT-5.6 Sol Pro is the same underlying model as [GPT-5.6 Sol](https://openrouter.ai/openai/gpt-5.6-sol), served with `reasoning.mode` set to `pro` for higher-quality responses on complex tasks. Learn more in OpenAI's docs: https://developers.openai.com/api/docs/guides/reasoning#reasoning-mode
Grok 4.6 is SpaceXAI's smartest model with frontier performance on coding, knowledge work, and STEM.
This model always redirects to the latest model in the MoonshotAI Kimi family.
Claude Opus 5.5 is Anthropic's flagship model for demanding reasoning, coding, and long-horizon agentic work, succeeding Claude Opus 5. It is particularly strong at multi-step changes in large codebases, code...
Claude Opus 5.5 is Anthropic's flagship model for demanding reasoning, coding, and long-horizon agentic work, succeeding Claude Opus 5. It is particularly strong at multi-step changes in large codebases, code...
Qwen3.8 Max is the flagship model in Alibaba's Qwen3.8 series, the general-availability successor to the Qwen3.8 Max Preview. It is a multimodal reasoning model intended for complex reasoning, visual understanding,...
Qwen3.8 2.4T A95B is an open-weight sparse mixture-of-experts model from Qwen and the open-weight variant of [Qwen3.8 Max](/qwen/qwen3.8-max), with 95 billion active parameters out of 2.4 trillion total. It is...
Muse Spark 1.2 is a reasoning model from Meta, designed for complex agentic tasks. It accepts text, images, video, audio, and PDF documents, returns text, and offers a 1M-token context...
Muse Spark 1.2 contributor tier is a reasoning model from Meta designed for developers who want to start building at an even lower cost. It’s meaningfully cheaper than Muse Spark...
This model always redirects to the latest model in the Claude Sonnet family.
Claude Sonnet 5.5 is Anthropic's Sonnet-class model for well-scoped everyday work, succeeding Claude Sonnet 5 as a direct upgrade. It is especially strong at building features, fixing bugs, and producing...