Claude Opus 4.6: Anthropic's Most Powerful AI Model Yet

Anthropic Raises the Bar with Claude Opus 4.6

Anthropic released Claude Opus 4.6 on February 5th, 2026, just three months after launching Opus 4.5, marking an aggressive development pace in the competitive AI landscape. The model is positioned as the company's most capable to date, with significant improvements across coding, financial analysis, legal research, and agentic workflows.

Benchmark Dominance

Opus 4.6 achieved a 65.4 percent score on Terminal-Bench 2.0, an evaluation of agentic coding capabilities, and an Elo rating of 1,606 on GDPval-AA, a benchmark measuring performance on real-world knowledge work tasks. That score places it roughly 150 points ahead of OpenAI's GPT-5.2, implying a win rate of about 70 percent in head-to-head comparisons. On ARC-AGI-2, the model scored 68.8 percent, a dramatic leap from Opus 4.5's 37.6 percent.

Enterprise-Focused Features

The model introduces a one-million-token context window in beta, matching what Anthropic's Sonnet models currently offer. In Claude Code, users can now deploy agent teams that split tasks across multiple coordinated agents working in parallel. Adaptive thinking allows the model to automatically extend its reasoning time for complex tasks, with developers able to choose between four effort levels.

Beyond Coding

Anthropic is expanding Claude's reach beyond developers. On the Finance Agent benchmark, Opus 4.6 achieved 60.7 percent, a 5.47 percentage point improvement over its predecessor. The company also launched Claude in PowerPoint as a research preview, allowing users to build and edit presentations directly within Microsoft software. Pricing remains at five dollars per million input tokens and twenty-five dollars per million output tokens, though the model uses 30 to 60 percent more tokens than Opus 4.5 in adaptive thinking mode.