Anthropic introduced Claude Haiku 5.5 on October 7, describing the model as an option for high-volume, cost-sensitive AI workloads. The company highlights short tasks such as summarization, classification, database queries and background subagent work, where response time and per-request cost can matter as much as maximum model capability.

According to Anthropic, Haiku 5.5 is its fastest small model at standard speed. The company says its average cost per task is approximately 75% lower than Haiku 4.5, with the exact savings depending on prompt size and token usage. Anthropic's published pricing includes different rates for prompts above and below 100,000 tokens, so organizations should check their actual workloads rather than assume one headline saving applies to every request.

Anthropic says the model is available through the Claude Platform and cloud partners including AWS, Google Cloud and Microsoft Azure. Its API model identifier is claude-haiku-5-5.

The release also accompanies a reduction in Claude Sonnet 5.5 cache-read pricing, which Anthropic says lowers costs for many agentic workloads. For developers selecting models, the practical choice remains workload-specific: a smaller, faster model may suit repetitive support and classification tasks while larger models remain useful for more demanding work. Detailed pricing and evaluations are available from Anthropic's original release.