Claude Haiku 5.5 is the fastest model in Anthropic’s Claude 5.5 family and the successor to Claude 4.5 Haiku. It scores 48.4% on CursorBench at max effort, on the same cost-performance curve as Sonnet 5.5 and Opus 5.5, at a tenth of Haiku 4.5’s per-token price for requests up to 100k input tokens. Add it to the model picker from Cursor Settings > Models.
Strengths¶
- Low cost for high-volume work. Requests up to 100k input tokens bill at $0.10/M input and $0.50/M output.
- The first Haiku with effort levels. With thinking on, its CursorBench score climbs from 30.9% at low effort to 48.4% at max.
- Thinking can be turned off at low, medium, and high effort for faster, cheaper responses.
- Supports context windows up to 1M tokens.
Limitations¶
- Requests above 100k input tokens bill at 5x the standard rates. Long agent sessions with large contexts cost more per token once they cross that line.
- Sonnet 5.5 and Opus 5.5 score higher on CursorBench for harder tasks.
- Turning thinking off lowers quality. On CursorBench it scores 22.1% to 26.2% with thinking off, compared with 30.9% to 42.3% with thinking on at the same effort levels.
Tools¶
Haiku 5.5 has access to all agent tools when used with Cursor including:
Learn more about how tools work and tool calling fundamentals.
Pricing¶
Haiku 5.5 is a third-party model, so it isn’t part of the usage included with Cursor plans. On The X Subscription it’s billed at the rates below from your prepaid credit. On legacy Cursor plans it draws from the Other Models pool. All prices are per million tokens.
Requests up to 100k input tokens bill at $0.10/M input and $0.50/M output. Prompt-cache writes are $0.125/M and cache reads are $0.01/M.
Requests above 100k input tokens (long context) bill at 5x those rates: $0.50/M input, $0.625/M cache write, $0.05/M cache read, and $2.50/M output.
A thinking variant is available for deeper reasoning.
