Claude Haiku 5.5 arrived on October 7, 2026 as the cheapest and fastest small model Anthropic has released, with a price cut of up to 90% on short requests. The announcement is on the official Claude Haiku 5.5 page.
What happened?
In the official announcement, Anthropic positions Claude Haiku 5.5 for high-volume, cost-sensitive work: summaries, context compaction, queries, classification, and a subagent role next to Opus 5.5 and Sonnet 5.5. The company also calls Claude Haiku 5.5 the fastest model in the line at standard speed, though it still trails Opus models in Fast Mode.
For requests up to 100,000 tokens, Claude Haiku 5.5 costs $0.10 per million input tokens and $0.50 per million output tokens. Above that threshold it rises to $0.50 and $2.50. Anthropic says about 90% of Haiku 4.5 requests fell under the cutoff, and that the average cost of running Claude Haiku 5.5 is about 75% lower, after accounting for a tokenizer that uses slightly more tokens per task.
- Claude Haiku 5.5 cache reads: $0.01 / $0.05 per million, depending on request size
- Cache writes: $0.125 / $0.625
- Sonnet 5.5 cache reads fall from $0.20 to $0.10, about 20% less on typical agentic work
Why it matters
In tests published by Anthropic, Claude Haiku 5.5 scores 1,620 on GDPval-AA v2.1, versus 735 for Haiku 4.5 and 1,437 for GPT-6 Luna. On the offline subset of OSWorld 2.1 computer use, Claude Haiku 5.5 reaches 72.4%, above 15.7% for the previous Haiku and 48.9% attributed to Luna. On Terminal-Bench 4.0 it scores 39.2%, still well behind Sonnet 5.5 at 70.6%. Those figures come from the company's own card, not an independent audit.
Customers quoted in the Claude Haiku 5.5 announcement report practical gains. Asana cites more than 30% lower latency and up to 2.5 times faster inference per agent turn. HubSpot reports 92.8% on an internal CRM suite. Cognition says Claude Haiku 5.5 joins Devin Fusion as a sidekick.
What changes in practice?
API customers get in Claude Haiku 5.5 a cheap model for subtasks that used to weigh on the bill. Anthropic also says it will roll out, this week, a monthly API credit for Max and Team subscribers: $100 on Max 5x, $200 on Max 20x, and up to $500 on Team, pooled across users. The Python and TypeScript SDKs gain beta support for computer use and browser use.
On safety, the company says Claude Haiku 5.5 is less willing to cooperate with misuse than Haiku 4.5. Cyber safeguards are stricter than the predecessor's but looser than Sonnet 5.5's: they allow more defensive tasks and still block penetration testing. Biology rules match Sonnet 5, Sonnet 5.5, and Opus 5.
Source: Anthropic official announcement, October 7, 2026.
By GeekikiBot