Anthropic launches Claude Haiku 5.5 for high-volume, low-latency work

Anthropic released Claude Haiku 5.5 on October 7, positioning it as a small, fast model for high-volume tasks such as classification, summarization, database queries, live support and narrowly scoped subagent work. It is available now as claude-haiku-5-5 through the Claude Platform and through Amazon Web Services, Google Cloud and Microsoft Azure. The model remains proprietary and is offered as a hosted service rather than downloadable weights.

Pricing now varies by prompt length. For prompts up to 100,000 tokens, Anthropic lists $0.10 per million input tokens and $0.50 per million output tokens. Above that threshold, prices rise to $0.50 and $2.50, respectively. Haiku 5.5 is also the first Haiku model with an adjustable effort setting, letting developers trade cost and latency against capability.

Anthropic reports large gains over Haiku 4.5: 72.4% versus 15.7% on the offline subset of OSWorld 2.1, 39.2% versus 0% on Terminal-Bench 4.0, and 45.9% versus 10.2% on Humanity’s Last Exam without tools. It also reports a GDPval-AA v2.1 score of 1620, compared with 1437 for GPT-6 Luna. These are release-day, vendor-run results, and several use specific settings or subsets. They should not be treated as proof of general superiority without independent replication.

The launch targets scale and responsiveness rather than replacing Anthropic’s larger models. Anthropic explicitly says Sonnet 5.5 and Opus 5.5 remain better suited to complex agentic coding. It also applies tighter cyber safeguards than Haiku 4.5 and the same biology safeguards used for recent larger Claude models.

Sources: Anthropic launch announcement · Claude Haiku product page · Unsplash image license