Claude Sonnet 4.6

Claude Sonnet 4.6, Anthropic's mid-tier model, is defined by a single number: 72.1% on OSWorld-Verified — exactly the human baseline. That parity is its strongest claim and its biggest risk. The model deploys Chain-of-Thought Prompting and Constitutional AI, but those techniques are now table stakes. The real pressure comes from within: Claude Sonnet 5 already beats Opus 4.8 on knowledge work at lower cost, making Sonnet 4.6 look like a stepping stone. Anthropic is racing to ship Claude Lite for cost-conscious enterprises, but Sonnet 4.6 is caught in the middle — not cheap enough to be Lite, not capable enough to be Sonnet 5. Mention volume has collapsed to just 2 in the last 30 days. The model powers Claude Code, but as the ecosystem shifts toward lighter, cheaper, or smarter alternatives, Sonnet 4.6 risks becoming the forgotten middle child.

• ·Scores exactly 72.1% on OSWorld-Verified, matching the human baseline
• ·Competes internally with Claude Sonnet 5, which outperforms Opus 4.8 at lower cost
• ·Deploys Chain-of-Thought and Constitutional AI, now commoditized techniques
• ·Mention volume dropped to 2 in last 30 days, signaling fading relevance
• ·Powers Claude Code but faces squeeze from new Claude Lite and Sonnet 5