Overview
Released on October 7, 2026, Claude Haiku 5.5 is Anthropic’s small model for high-volume, cost-sensitive, latency-sensitive work. Anthropic describes it as its fastest and most capable small model at launch. (anthropic.com)
Capabilities and uses
- Accepts text and images and produces text; supports adaptive thinking with an adjustable effort parameter.
- Provides a 1 million-token context window and up to 128,000 output tokens. The Message Batches API supports up to 300,000 output tokens with a beta header.
- Suitable for summarization, classification, extraction, routing, database queries, live customer support, browser automation, focused coding, and subagent work alongside larger models.
- Available through the Claude API, Claude Code, Claude apps, and supported cloud platforms. (platform.claude.com)
Limitations
- Knowledge and training-data cutoff: June 2026. Outputs can be inaccurate and should be reviewed, particularly for consequential tasks.
- Migration from Haiku 4.5 can require code changes: manual thinking budgets, sampling parameters such as
temperature,top_p, andtop_k, and assistant-message prefilling can produce errors. The newer tokenizer also produces approximately 30% more tokens for the same text than Haiku 4.5. - Anthropic reports safety improvements over Haiku 4.5, but these do not eliminate model or misuse risks. (platform.claude.com)
API pricing
Currency: USD. Unit: per 1 million tokens. The following are Claude API prices with default global routing, checked November 15, 2026. Each request’s prompt length includes input tokens, cache reads, and cache writes. When the prompt exceeds 100,000 tokens, the higher tier applies to that request. (platform.claude.com)
| Charge | Prompt up to 100,000 tokens | Prompt over 100,000 tokens |
|---|---|---|
| Input | $0.10 | $0.50 |
| Output | $0.50 | $2.50 |
| 5-minute cache write | $0.125 | $0.625 |
| 1-hour cache write | $0.20 | $1.00 |
| Cache read | $0.01 | $0.05 |
| Batch API input | $0.05 | $0.25 |
| Batch API output | $0.25 | $1.25 |
The Batch API is asynchronous and discounts input and output by 50%. US-only inference on the first-party Claude API applies a 1.1× multiplier to token pricing. Partner-operated platforms can have independent regional prices; server-side tools may also incur additional usage charges. (platform.claude.com)