Skip to content

feat: support Anthropic adaptive thinking#7356

Open
rabi wants to merge 1 commit intoblock:mainfrom
rabi:adaptive_thinking
Open

feat: support Anthropic adaptive thinking#7356
rabi wants to merge 1 commit intoblock:mainfrom
rabi:adaptive_thinking

Conversation

@rabi
Copy link
Contributor

@rabi rabi commented Feb 19, 2026

Summary

Claude Opus 4.6 and Sonnet 4.6 support adaptive thinking where the model dynamically determines when and how much to think, replacing the fixed budget_tokens approach. This is especially important for agentic workflows since interleaved thinking (thinking between tool calls) requires adaptive mode on these models.

Config: claude_thinking_type (adaptive|enabled|disabled), claude_thinking_effort (low|medium|high|max). Defaults to adaptive for 4.6 models, preserves existing env var gating for older models.

Type of Change

  • Feature
  • Bug fix
  • Refactor / Code quality
  • Performance improvement
  • Documentation
  • Tests
  • Security fix
  • Build / Release
  • Other (specify below)

AI Assistance

  • This PR was created or reviewed with AI assistance

Testing

Tested locally.

Closes: #7293

Add adaptive thinking support for Claude 4.6 models (Opus, Sonnet)
with three thinking modes: adaptive, enabled (budget-based), and
disabled.

Change-Id: I33dbf77bf50ad78e909b3600705b706999bf1ebf
Signed-off-by: rabi <ramishra@redhat.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Support Anthropic adaptive thinking API (replace deprecated budget_tokens)

1 participant

Comments