Introducing Claude Haiku 5.5

Open original

Anthropic released Claude Haiku 5.5, a small language model priced 75% lower than its predecessor while delivering significantly better performance on tasks like summarization, customer support, and coding. The company also cut Sonnet 5.5 cache read prices in half and added monthly API credits for paid subscribers.

  • Haiku 5.5 costs $0.10 per million input tokens (vs. $1.00 for Haiku 4.5) and outperforms Haiku 4.5 across benchmarks—scoring 72.4% on OSWorld computer use tasks compared to 15.7% for the older model.
  • The model includes an adjustable effort setting that lets users trade off speed for accuracy, and is positioned as ideal for high-volume, cost-sensitive work like batch summarization, database queries, and subagent tasks in larger AI systems.
  • Sonnet 5.5 cache reads now cost 50% less ($0.10 instead of $0.20 per million tokens), reducing agentic task costs by ~20%, while Max and Team subscribers get monthly API credits ($100–$500) to experiment with building agents and applications.

Anthropic just released Claude Haiku 5.5, a cheaper and faster version of their small language model. The headline numbers: it costs about 75% less than Haiku 4.5, and for requests under 100,000 tokens (which make up 90% of typical usage), it's 90% cheaper. The model handles repetitive work well—summaries, database queries, classification, customer support—and works fast enough for live applications. On benchmarks, Haiku 5.5 shows meaningful jumps over its predecessor. On knowledge work tests, it scores around 1600 versus 735 for Haiku 4.5. For computer use tasks, it hits 72.4% accuracy compared to 15.7% before. It still trails their larger models (Sonnet 5.5 and Opus 5.5) on complex agentic coding, but that's not its job—it's built for tasks that were previously too expensive to automate at scale. The pricing reflects the positioning. Input tokens cost $0.10 per million for prompts under 100k tokens, dropping to $0.50 for longer ones. Output tokens run $0.50 and $2.50 respectively. For comparison, Haiku 4.5 cost $1.00 and $5.00. Haiku 5.5 also supports an adjustable effort setting, letting users trade off between cost and accuracy depending on the task. Anthropic also cut Sonnet 5.5's cache read prices in half, which reduces agentic task costs by around 20% since cache reads dominate token usage in those workflows. On safety, Haiku 5.5 shows fewer instances of misaligned behavior and lower willingness to cooperate with misuse compared to Haiku 4.5. Its cybersecurity safeguards are tighter than the previous version but looser than Sonnet 5.5—it allows defensive security work but blocks penetration testing. Biology safeguards match those of the larger models, allowing research questions while restricting potentially harmful requests. The model launches today across AWS, Google Cloud, and Azure. Anthropic also added $100-500 monthly API credits for Max and Team subscribers to experiment with building agents and applications.

Ask about this article