Introducing Claude Opus 4.6

• First Opus-class model with 1M token context window (beta).
  • 76% on MRCR v2 8-needle 1M variant; Sonnet 4.5 scored 18.5%.
  • Premium 1M context: $10/$37.50 per MTok above 200k tokens.
• Beats GPT-5.2 by ~144 Elo on GDPval-AA agentic eval.
  • Exceeds Opus 4.5 by 190 points on same benchmark.
• Leads Terminal-Bench 2.0, Humanity’s Last Exam, BrowseComp.
• Nearly 2× Opus 4.5 on computational and structural biology tests.
• Lowest over-refusal rate of any recent Claude model.
• Base pricing unchanged: $5/$25 per million input/output tokens.
• Agent teams in Claude Code + context compaction ship as beta.

· ** · Read on anthropic.com