> ## Documentation Index
> Fetch the complete documentation index at: https://docs.apiyi.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Claude thinking and effort guide refreshed

> Rewritten for Opus 5.5 / Sonnet 5.5 / Haiku 5.5 / Fable 5.1, with thinking defaults, on/off support, the 128,000 output cap, and corrected Claude parameters in the empty-response FAQ.

**2026/11/9 10:40 (UTC+8)** · Docs Update · Claude

📖 **Claude thinking and effort guide rewritten for 5.x**

The old page still described the Opus 4.8 / 4.7 generation. It now covers the current lineup: Opus 5.5 / Sonnet 5.5 / Haiku 5.5 think by default, so "omit `thinking` and it won't think" no longer applies. Opus 5.5 and Fable 5.1 cannot disable thinking; Sonnet 5.5 uses `between_tools`; Haiku 5.5 and Opus 5 can disable it when effort is `high` or lower.

It also documents effort defaults (only Opus 5.5 / Haiku 5.5 default to `medium`; the rest default to `high`), the 128,000 max output, thinking tokens counting toward `max_tokens`, `display` and streaming events, common 400s, and official doc links.

The Claude rows in the reasoning-model empty-response FAQ were corrected too: the effort parameter is now `output_config.effort`, and `budget_tokens` is gone on 4.7 and newer.

📖 [Claude thinking guide](https://docs.apiyi.com/en/api-capabilities/claude-effort-thinking.md) · 📖 [Empty-response FAQ](https://docs.apiyi.com/en/faq/reasoning-model-empty-response.md)

***

← [Back to live updates](https://docs.apiyi.com/en/live.md) · 📚 [Monthly archive](https://docs.apiyi.com/en/live/archive.md)


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.