First model to exceed 80% on SWE-bench Verified. 200K token context. Also released as Claude 4.5 Sonnet (September 2025, 1M context, AA 37) and Claude 4.5 Haiku (October 2025, AA 31). Architecture undisclosed.

AA Intelligence Index: 43. Claude 4.5 Opus was the highest-ranked non-reasoning model at the time of release, with particular strength in coding, writing, and multi-step agentic tasks. Proprietary.

Model Details

Context window 200,000

Variants

Name Parameters Notes
Claude 4.5 Sonnet 1M context, AA index 37
Claude 4.5 Haiku 200K context, AA index 31
Claude 4.5 Opus 200K context, AA index 43
frontiercodingreasoning

Related