Anthropic released Claude Haiku 5.5 on October 7. The company positions it as a small model for high-volume, cost-sensitive work such as summarisation, classification, routing, live support and narrowly defined subagent tasks. It is available through Claude products, the Claude Platform and major cloud providers.
For prompts up to 100,000 tokens, Anthropic lists API prices of US$0.10 per million input tokens and US$0.50 per million output tokens. Above 100,000 prompt tokens, the listed rates rise to US$0.50 for input and US$2.50 for output. Cache writes and reads have separate rates, so the final bill depends on prompt length, output length, caching and repeated calls.
Anthropic says Haiku 5.5 costs about 75% less per task on average than Haiku 4.5, after accounting for pricing and token-use changes. That is a company estimate based on its workload mix, not an independent guarantee for every application. The newer tokenizer can also use slightly more tokens for the same task, so developers should test real workloads.
The model is the first Haiku release with an adjustable effort setting, allowing developers to trade cost and speed against more reasoning. Anthropic also presents it as suitable for computer and browser use, but those capabilities still depend on the surrounding application, permissions and safeguards; releasing a model does not itself give it access to a user’s computer or accounts.
The launch includes company benchmark results showing large gains over Haiku 4.5, but benchmark scores do not establish identical performance in production. Anthropic says Haiku 5.5 has stricter cyber safeguards than Haiku 4.5 and biology safeguards aligned with its larger recent models. Buyers should evaluate accuracy, latency, safety and total cost on their own tasks before replacing an existing model.