AI news · October 8, 2026
Claude Haiku 5.5 brings cheaper agent work to a high-volume model
Haiku 5.5 costs $0.10 per million input tokens and $0.50 per million output tokens up to 100k context. Its lower price makes it a serious first pass for repetitive agent work.
- input/output price per million tokens
- $0.10/$0.50
- OSWorld score
- 72.4%
- lower average cost than Haiku 4.5
- 75%
Anthropic released Claude Haiku 5.5 on October 7 as a high-volume model for coding, browser work, and computer use. The published price is $0.10 per million input tokens and $0.50 per million output tokens for inputs up to 100k tokens, with cache reads at $0.01 per million. Anthropic says average cost is 75% lower than Haiku 4.5. Its own tests report 72.4% on OSWorld and 39.2% on Terminal-Bench, which puts it above earlier Haiku versions but below Sonnet 5.5 on those tasks.
The practical use is task routing: send easy extraction, support, code review, and repetitive browser jobs to Haiku, then escalate only the hard cases. The model is available through the API and major cloud providers, with browser and computer-use tools in beta. Builders should measure error rate, retries, and total task cost instead of choosing on token price alone. A small model that needs two retries can cost more than a larger first pass.
What you can do with it
Try Haiku 5.5 in the Anthropic API or a listed cloud provider. Use a fixed set of support, extraction, coding, and browser tasks. Record success rate, retries, time to completion, and total token cost against the model you use today before changing your default.
Our take
This is the strongest small-model release in the window because the price change can alter system design. It should handle routine work first, but the published benchmark gap to Sonnet 5.5 means teams still need a clear escalation path for long or fragile tasks.
Links Try today HN discussion
Source: Anthropic ↗ — Made With Models writes the brief; the reporting is theirs.