Particle.news

Download on the App Store

Anthropic Releases Claude Haiku 4.5 With Near-Frontier Coding at Lower Cost

The lightweight model is now broadly available with low pricing, an ASL-2 safety designation, and a design tailored for real-time, multi-agent deployments.

Overview

  • Haiku 4.5 is available today across Anthropic apps and Claude Code, the API, Amazon Bedrock, and Google Cloud Vertex AI, priced at $1 per million input tokens and $5 per million output tokens.
  • Anthropic says the model delivers coding performance comparable to Sonnet 4 at roughly one-third the cost and more than twice the speed, and it reportedly surpasses Sonnet 4 on certain computer-use tasks.
  • In company tests, Haiku 4.5 scored about 73% on SWE-Bench Verified and around 41% on Terminal-Bench, placing it on par with Sonnet 4 and close to leading models on those benchmarks.
  • The release is positioned for real-time, low-latency work and multi-agent workflows, with Sonnet 4.5 planning tasks and multiple Haiku 4.5 instances executing subtasks in parallel.
  • Anthropic reports lower rates of misaligned behavior than prior models and assigns Haiku 4.5 an internal ASL-2 safety level, and the company says it is rolling out as the default for free Claude plans.