Overview
Anthropic positions Haiku 5.5 for high-volume developer workflows, including classification, summaries and tool-using agents.
Release and Availability
Announced October 7, 2026. Anthropic lists access through its API and major cloud partners. Provider access and regional eligibility can differ.
Official Specifications
| Detail | Official reference / unknown |
|---|---|
| Release | October 7, 2026 · Source reviewed |
| API model ID | claude-haiku-5-5 · Source reviewed |
| Context window | 1M tokens · Source reviewed |
| Maximum output | 128K tokens · Source reviewed |
| Thinking | Adaptive; medium default · Source reviewed |
| Parameters | Not publicly confirmed · Source reviewed |
| Downloadable weights | Not publicly confirmed · Source reviewed |
API Access
Use claude-haiku-5-5 in the Claude API. Keep credentials on the server, validate tool arguments and evaluate the default thinking setting before changing it.
API Pricing Summary
Standard uncached USD rates are $0.10 input / $0.50 output per million tokens for requests with up to 100K input tokens. Above that threshold, the entire request uses $0.50 / $2.50. Monthly volume does not determine this tier.
Benchmarks Summary
Published evaluation results depend on the harness and reasoning settings. Our benchmark guide separates Anthropic comparisons from independent Artificial Analysis results. AI Model Brief has not run inference benchmarks.
Use Cases
- Classify and route incoming work.
- Compact long agent conversations.
- Evaluate coding assistants and browser workflows.
Limitations and Unknowns
Parameter count and a downloadable weight license: Not publicly confirmed. API access does not establish open weights. Validate factual answers and tool actions, and measure latency on your own workload.
Frequently asked questions
Is Claude Haiku 5.5 an official model?
Yes. Anthropic documents its launch and the claude-haiku-5-5 API ID in the official sources linked here.
Does the advertised token rate apply to every request?
No. Standard rates change when the input to one request exceeds the threshold. Cache, batch and other billing terms also affect cost; use the pricing guide.
Go straight to the source
Reviewed Oct 8, 2026. Official statements, third-party results and arithmetic estimates are attributed separately. Unknowns reflect the sources reviewed on this date. No inference benchmarks were run by this site.