Best official AI provider docs and model catalogs

Model names, aliases, context windows, pricing, tool support, and recommended usage can change quickly. Official docs and model catalogs are the first place to check before relying on a course, tweet, or old benchmark.

Bookmark OpenAI, Anthropic, Gemini, Llama, Mistral, Cohere, DeepSeek, Qwen, xAI, Perplexity, Together, and OpenRouter docs if your work depends on model behavior. As of Wednesday, July 22, 2026, that means checking pages that already show the current families and migrations, plus the operational docs that affect real costs and latency: GPT-5.6 model guidance plus compare-models, the Responses API migration guide, conversation state, Agent Builder, Codex best practices, Batch, Flex, and Prompt Caching in OpenAI's docs; Anthropic Academy plus Claude overview, the current Claude lineup, Managed Agents, Advisor tool, Prompt Caching, Batch Processing, and release notes; Gemini model tables plus Agents Overview, Batch API, context caching, pricing, the GA Interactions API, URL Context, and deprecations in Google's docs; Devstral, Magistral, Voxtral, and OCR updates in Mistral's docs; and OpenRouter's large model directory plus Fusion Router for cross-provider discovery. Secondary summaries are useful only after the primary source is checked.

A model catalog should not make the final decision for you. It should narrow the candidates by modality, context, tool support, speed, price, hosting route, and data constraints.

Once the list is short, test the models on your own prompts and workflows. That is the only way to know whether a model is right for your coding agent, RAG system, product feature, or research assistant.