AI BriefingClaude Haiku 4.5
Key point
Anthropic has released a new small model, Claude Haiku 4.5, offering excellent cost-efficiency and speed.
Details
Anthropic has unveiled its new small model, Claude Haiku 4.5. This model delivers coding performance similar to the existing Claude Sonnet 4, while costing one-third as much and running more than twice as fast.
In particular, it outperforms Claude Sonnet 4 on certain tasks such as computer use. It is optimized for low-latency tasks like real-time chat assistants, customer service agents, and pair programming.
Key features are as follows:
- Cost efficiency: Delivers Sonnet 4-level performance at 1/3 the price
- Speed: Response speed more than twice as fast as the previous model
- Collaborative model: A structure is possible where Claude Sonnet 4.5 plans complex problems while multiple instances of Haiku 4.5 carry out subtasks in parallel
There have also been advances on the safety front. Claude Haiku 4.5 shows a higher level of alignment than its predecessor, Haiku 3.5, and recorded a lower rate of misaligned behavior than Claude Sonnet 4.5 and Opus 4.1 in automated evaluations, making it the safest model to date. Accordingly, it has been released at the ASL-2 level.
It is currently available via the Claude API, priced at $1 per million input tokens and $5 per million output tokens. It is also available on Amazon Bedrock and Google Cloud Vertex AI.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.