Claude Opus 4.6

Claude Opus 4.6 shipped on February 5, 2026 as the first Opus-class model with a 1M-token context window (in beta at launch). Its headline feature is adaptive thinking: the model calibrates reasoning depth to task complexity across four effort levels (low, medium, high, max). At release it set state-of-the-art scores on Terminal-Bench 2.0 and Humanity's Last Exam, with 80.8% on SWE-bench Verified and 91.3% on GPQA Diamond at max effort, alongside agent-teams in research preview.

Pricing held at $5/$25 per million tokens, with premium $10/$37.50 rates for prompts over 200K tokens. The caveat in late 2026 is generational: Opus 4.7, 4.8, and the 5.x family have since shipped, so 4.6 is now a previous-generation pin for workflows tuned against it rather than a starting point for new projects.