Anthropic has released Claude Haiku 5.5, the latest version of its smallest, fastest model line. The company calls it the cheapest and most capable small model it has shipped, and the numbers back that framing up.
The context window jumps to one million tokens, up from 200,000 on the previous Haiku, with output capacity rising to 128,000 tokens from 64,000. That is enough to hold large codebases, long documents or extended conversation history in a single pass, which previously meant stepping up to a bigger, pricier model.
Pricing is the other headline. For prompts up to 100,000 tokens, input costs $0.10 per million tokens and output costs $0.50 per million. Cached reads drop to $0.01 per million. Longer prompts, over 100,000 tokens, move to $0.50 input and $2.50 output per million. Even at the higher tier, Haiku 5.5 undercuts most mid-sized models by a wide margin.
On benchmarks, Anthropic reports 72.4 per cent on the offline subset of OSWorld 2.1, a computer-use test, and 45.9 per cent on Humanity's Last Exam without tool access, rising to 57.4 per cent with tools. On Terminal-Bench 4.0, a coding and command-line task, it scores 39.2 per cent.
Haiku 5.5 is live now on Anthropic's own Claude Platform and through Amazon Web Services, Google Cloud and Microsoft Azure. For teams running high-volume workloads, such as classification, extraction or routing tasks, where a long context and low per-token cost matter more than frontier reasoning, this is the model Anthropic is positioning as the default choice.
Source: MarkTechPost
