Anthropic Releases Claude Haiku 5.5, Its Fastest Model Yet

Anthropic Releases Claude Haiku 5.5, Its Fastest Model Yet

Anthropic on Thursday introduced Claude Haiku 5.5 and called it the fastest model it has released so far. The company is positioning it for developers who run large numbers of requests and need to keep costs down. Typical uses include summarisation, classification, database queries, and context compaction. It is also aimed at browser-based apps and coding work done by smaller helper agents.

Anthropic says Haiku 5.5 costs about 75 percent less to run than Haiku 4.5. It is also the first Haiku model with an adjustable effort setting.

Pricing is built around prompts of up to 100,000 tokens. The company says that range covered about 90 percent of requests to the previous Haiku model. Below that limit, input costs $0.10 per million tokens (about Rs. 9.67). Output costs $0.50 per million tokens (about Rs. 48.38). Cache reads are $0.01 per million tokens (about Rs. 0.96), and cache writes are $0.125 per million tokens (about Rs. 12.57).

Above 100,000 tokens, rates rise. Input becomes $0.50 per million tokens (about Rs. 48.38). Output becomes $2.50 per million tokens (about Rs. 241). Cache reads rise to $0.05 per million tokens (about Rs. 4.83), and cache writes rise to $0.625 per million tokens (about Rs. 60.95).

The lower price is meant for jobs where speed and running cost matter more than the peak ability of Anthropic’s larger models. Haiku 5.5 can also work as a subagent on coding tasks next to Claude Opus 5.5 and Sonnet 5.5. Anthropic points to live customer support and browser use as fits for the model.

The new effort control lets users choose whether a job should favour lower cost or stronger reasoning. For developers, Anthropic plans beta support for computer use and browser use in its Claude Python and TypeScript software kits. The company says the mix of speed, ability, and lower cost makes the model better suited to those tasks.

On safety, Anthropic says Haiku 5.5 scores better than Haiku 4.5 on nearly all of its alignment tests. It is also described as less willing to assist misuse.

The model is available on Anthropic’s own platforms and through Amazon Web Services, Google Cloud, and Microsoft Azure. Developers can call it on the Claude Platform with the model name claude-haiku-5-5.

Disclaimer: This article is a general news summary based on the company’s public announcement. Prices are approximate conversions and may change. Performance and safety claims are as stated by the maker and have not been independently verified here. This is not a product recommendation.