Claude Opus 5.5 has shipped, leading SimpleBench with an 88.4% score and generating significant community attention for its ability to produce high-quality explainer videos.

  • On vision evaluations, it is ranked as Anthropic's best vision model to date, outperforming Fable 5 and GPT-6 Sol while costing about 60% less than Fable 5.1.
  • Reasoning performance on Terminal-Bench-Science peaks at 62% with xhigh effort, though it drops to 59% at max effort due to imposed minimum budgets.
  • It leads the Terminal-Bench-Science leaderboard alongside GPT-6 Astra, surpassing Fable 5.1 by approximately 20 points.
  • The $200 Claude Code plan is reported to beat Codex, while Astra remains the preferred model for review and audit tasks.

The release establishes Opus 5.5 as a top-tier frontier model with strong vision capabilities and cost efficiency compared to competitors.