Since OpenAI and friends refuse to give us a max_ctx param in /models, here's the current context window, input token and output token limits for OpenAI (API), Anthropic, Qwen, Deepseek, llama, Phi, Gemini and Mistral - View it on GitHub