claude-haiku-5-5. Check the model plaza for current availability and pricing.
Highlights
- Frequent tasks: handle repeated requests, document summaries, and narrowly scoped supporting work.
- Adaptive thinking: enabled by default; adjust thinking depth with
effort. The default level ismedium. - Long context and vision: Anthropic lists a 1M-token context window, up to 128K output tokens, and text and image input with text output.
Before switching from Haiku 4.5
For direct Messages API calls, review the official migration guide. Replace the oldbudget_tokens configuration with adaptive thinking, omit temperature, top_p, and top_k, and remeasure token usage and max_tokens limits. The new tokenizer and thinking behavior affect existing token budgets.
Connect it
- Create a BetterToken API Key or use an existing key.
- In Claude Code or an Anthropic-protocol client, set the Base URL to
https://bettertoken.ai. - Set the Model ID to
claude-haiku-5-5, without a date suffix. - Test quality, latency, and token usage on a representative task before switching an existing workflow.

