Anthropic has introduced Claude Haiku 5.5, describing it as the cheapest, fastest and most capable small model it has ever released. According to the company’s announcement, Haiku 5.5 is designed for high‑volume, cost‑sensitive workloads such as summaries, compactions, database queries and classification requests. It also pairs well with the Opus 5.5 and Sonnet 5.5 models as a subagent on coding tasks and is said to be the fastest model to date, making it suitable for speed‑sensitive uses like live customer support and browser automation.
Pricing details show a significant reduction compared with the previous Haiku 4.5. Anthropic states that, on average, Haiku 5.5 now costs around 75% less to run. The pricing table indicates that for prompts up to 100,000 tokens the cost is $0.10 per million input tokens and $0.50 per million output tokens, while for longer prompts the rates rise to $0.50 and $2.50 respectively. For cache reads and writes the model charges $0.01/$0.05 and $0.125/$0.625 per million tokens, far below the $0.10/$1.25 and $0.10/$2.50 rates of Haiku 4.5. Footnote 2 notes that Haiku 5.5 is priced 90% lower than Haiku 4.5 for requests under 100,000 tokens and 50% lower for longer requests, with the average 75% figure reflecting the token‑usage shift caused by an updated tokenizer.
Performance figures from internal benchmarks are presented in the announcement. On knowledge‑work tests (GDPval‑AA v2.1 and AA‑Briefcase v1.1) Haiku 5.5 scores 1620 and 1578 points, outperforming Haiku 4.5’s 735 and 614 and approaching Sonnet 5.5’s 1840 and 1824. In the Computer use OSWorld 2.1 offline subset, Haiku 5.5 achieves 72.4% versus 15.7% for Haiku 4.5 and 48.9% for GPT‑6 Luna. On multidisciplinary reasoning (Humanity’s Last Exam) it reaches 45.9% without tools and 57.4% with tools, compared with 10.2%/18.7% for Haiku 4.5. Agentic coding benchmarks show 39.2% on Terminal‑Bench 4.0 and 46.4% on FrontierCode 1.1, well ahead of Haiku 4.5’s 0.0% and comparable to Sonnet 5.5’s 70.6% and 52.1%.
Haiku 5.5 is the first Haiku‑class model to include an adjustable effort setting, letting users trade off cost against intelligence. Early customer feedback cited in the release says results align with the advertised performance and cost improvements.
Availability is broad: the model is now live on Amazon Web Services, Google Cloud and Microsoft Azure, and can be accessed via the Claude Platform using the identifier claude‑haiku‑5‑5. Anthropic also announced a halving of Sonnet 5.5 cache‑read prices, a new monthly API credit for Max and Team subscribers, and beta updates to the Claude Python and TypeScript SDKs that add computer‑use and browser‑use support, noting Haiku 5.5’s speed, capability and low price make it a strong fit for those features.