Anthropic's small model of the Claude 4.5 generation, released 15 October 2025 and backfilled here. The announcement's pitch is that it delivers similar coding performance to Sonnet 4 from five months earlier at one third of the cost and more than twice the speed, with 73.3% on SWE-bench Verified, and that in Augment's agentic coding evaluation it reaches about 90% of Sonnet 4.5's performance while running four to five times faster. It supports extended thinking (benchmarks used a 128K thinking budget) and computer use, and is priced at $1 per million input tokens and $5 per million output on the Claude API, Amazon Bedrock, and Google Cloud Vertex AI, with a 200K context. Artificial Analysis scores the reasoning mode 17 on Intelligence Index v4.3 and the non-reasoning mode 15.