mirror of
https://github.com/browser-use/browser-use.git
synced 2026-09-14 19:59:47 +08:00
Chase Xu
993a4c1f5b
OpenAI's completion_tokens already includes reasoning_tokens as a subset, so adding them again was incorrectly inflating token counts by ~2x for all reasoning models (o1, o3, o4-mini, etc.). This aligns with other OpenAI-compatible providers (OpenRouter, Vercel, Cerebras, Groq) which use completion_tokens directly. Note: This is different from Google Gemini where thinking_tokens ARE reported separately and need to be added. Fixes #4065
Browser Use LLMs
We officially support the following LLMs:
-
OpenAI
-
Anthropic
-
Google
-
Groq
-
Ollama
-
DeepSeek
-
Mistral
Mistral specifics
Use ChatMistral with MISTRAL_API_KEY (and optional MISTRAL_BASE_URL). Structured outputs automatically strip unsupported JSON schema keywords (minLength, maxLength, pattern, format), and generation uses max_tokens plus the optional safe_prompt flag.
- Cerebras
Migrating from LangChain
Because of how we implemented the LLMs, we can technically support anything. If you want to use a LangChain model, you can use the ChatLangchain (NOT OFFICIALLY SUPPORTED) class.
You can find all the details in the LangChain example. We suggest you grab that code and use it as a reference.