Claude Sonnet 4.6 Gives You Flexibility

Anthropic has launched Claude Sonnet 4.6, a significant upgrade to its mid-tier model that blurs the line with its flagship Opus 4.6. The model is now available across all plans, including the free tier, Claude Code, the API, and major cloud platforms. This marks a substantial enhancement for non-paying users, who now get default access to Sonnet 4.6 with features previously locked behind paywalls: file creation, connectors, skills, and compaction.

Technically, Sonnet 4.6 represents a full skill upgrade across coding, computer use, long-context reasoning, agent planning, and design. It features a 1M token context window in beta. While it trails Opus 4.6 on some benchmarks like the ARC-AGI evaluation (86% vs. higher Opus scores), it surprisingly takes the lead on GDPval-AA. Performance analysis from Artificial Analysis shows Sonnet 4.6 now clusters with Opus 4.6 on the ELO vs. Cost curve, achieving a higher performance point while maintaining Sonnet 4.5's pricing of $3/$15 per million tokens for input/output. It shows a notable jump in Extended NYT connections to 58%, up from 49% in version 4.5.

The release reshapes Anthropic's model strategy. Sonnet 4.6 is positioned as a powerful, cost-effective alternative for scaling token use and for specific tasks like computer use where it may even be superior. It effectively replaces Haiku for many sub-agent workflows. The upgrade to the free tier democratizes access to near-frontier capabilities, particularly for coding and agentic tasks, though experts still recommend the $20/month plan for marginal quality gains in critical work. The model has sparked discussion about its 'warmer' tone, a noted change from Sonnet 4.5 that has received mixed reactions from the community.