# Claude Haiku 5.5 becomes the default fast model

> Claude Haiku 5.5 replaces Haiku 4.5 with a 1M-token window and an effort setting, becomes a default agent model, and leads Auto's Fast tier.

Oct 8, 2026. Topics: Models. https://www.generalinput.com/changelog/2026-10-08-claude-haiku-5-5

Anthropic released Claude Haiku 5.5 on October 7, and it reached General Input on October 8. It replaces Haiku 4.5 everywhere, and it's the first Haiku to be a default agent model rather than opt-in, so specialists can run on it. It appears in the model picker, the hosted API catalog, and the Subscription group when you connect a Claude plan through [Token Tunnel](/token-tunnel). See the [models page](/models).

### How it works

- **The model.** A 1M-token context window (Haiku 4.5 had 200K) and adaptive thinking with an effort setting that defaults to medium. Anthropic lists it at $0.10 input and $0.50 output per million tokens for prompts up to 100K tokens, and five times that above.
- **Saved selections move over.** Any workflow, specialist, or setting that named Haiku 4.5 now runs on Haiku 5.5.
- **Auto, hosted.** The Fast tier now tries Claude Haiku 5.5 first, then GPT-6 Luna, Gemini 3.8 Flash, and Grok 4.7.
- **Auto on your subscriptions.** When Auto can use a model you've connected through Token Tunnel, each tier now runs one class above its hosted counterpart, because subscription usage isn't billed per token: Fast uses Claude Sonnet 5.5 or GPT-6.1 Sol, Balanced uses Opus 5.5 or GPT-6 Astra.

### Also shipped

- A connected device that only serves Haiku 4.5 is no longer presented as Haiku 5.5.