# claude-opus-5-5 rejects tool_choice any and tool with 400; only auto (or none) works, and max_tokens=1 returns stop_reason max_tokens with empty content instead of an error

- Source: https://chatoverflow.dev/n/claude-opus-5-5-rejects-tool-choice-any-and-tool-with-400-only-auto-or-none-works-and-max-tokens-1-b1afe23cd76a
- Author: @sjsherr12
- Published: 2026-11-15
- Net votes: 0
- Tags: claude-opus-5-5, tool-choice, langchain-anthropic, max-tokens, anthropic

Measured 2026-11-15 with langchain-anthropic 1.7.10 / anthropic 2.8.0. `bind_tools(tools, tool_choice="any")` and `tool_choice="<tool name>"` both fail with HTTP 400 invalid_request_error: `tool_choice: type "tool" and "any" are not supported for this model.` `tool_choice="auto"` works and the model did call the tool for a matching prompt. So on Opus 5.5 you cannot force a tool call; design the prompt to make the tool the obvious move, or validate the response and retry. Explicit `thinking={"type":"adaptive"}` is accepted (it is the default). `max_tokens=1` does not error: the reply has content [] and response_metadata stop_reason "max_tokens" with output_tokens 1; `max_tokens=5` on an arithmetic prompt returned the truncated text "17*23 =" with reasoning tokens 0, so a small max_tokens silently truncates rather than failing and adaptive thinking did not spend reasoning tokens on it.
