Anthropic Releases Claude Sonnet 4.6 with Major Upgrades Across Coding, Computer Use, and Free Tier Access
Key Takeaways
- ▸Claude Sonnet 4.6 is now the default model for free tier users, offering Opus-level performance at Sonnet pricing ($3/$15 per million tokens)
- ▸The model shows major improvements in computer use capabilities, approaching human-level performance on complex real-world software tasks
- ▸Developers prefer Sonnet 4.6 over both its predecessor (70% preference) and the November 2025 flagship Opus 4.5 model (59% preference)
Summary
Anthropic has launched Claude Sonnet 4.6, its most capable Sonnet model to date, featuring significant improvements in coding, computer use, long-context reasoning, and agent planning. The model is now available across all plans, including the free tier, API, and major cloud platforms, with pricing remaining unchanged at $3/$15 per million tokens. Notably, Sonnet 4.6 now includes a 1M token context window in beta and has been made the default model for Free and Pro plan users on claude.ai and Claude Cowork.
The new model demonstrates substantial performance gains that rival the company's previous flagship Opus 4.5 model from November 2025. Early access developers prefer Sonnet 4.6 to its predecessor by wide margins, with 70% preferring it over Sonnet 4.5 and 59% choosing it over Opus 4.5 in Claude Code testing. Users report better instruction following, reduced "laziness" and overengineering, fewer hallucinations, and more consistent multi-step task completion.
Perhaps most notably, Sonnet 4.6 shows dramatic improvements in computer use capabilities, achieving human-level performance on tasks like navigating complex spreadsheets and filling multi-step web forms. Performance on OSWorld, the standard benchmark for AI computer use, has steadily improved over sixteen months of development. The model can interact with real software like Chrome, LibreOffice, and VS Code without special APIs, using virtual mouse and keyboard inputs. Anthropic has also enhanced the model's resistance to prompt injection attacks, a critical security concern for computer-using AI, with safety evaluations showing Sonnet 4.6 performs similarly to Opus 4.6 in this regard.
- Enhanced security features include improved resistance to prompt injection attacks, matching Opus 4.6 performance
- The model features a 1M token context window in beta and is available across all platforms including API and major cloud providers
Editorial Opinion
Anthropic's decision to make Sonnet 4.6 the default for free tier users represents a significant democratization of AI capabilities, effectively giving millions of users access to what performs comparably to their previous flagship model. The rapid improvement in computer use capabilities—from "experimental" in October 2024 to human-level performance in early 2026—suggests we're approaching a inflection point where AI agents can autonomously navigate legacy software systems. However, the acknowledgment of prompt injection vulnerabilities highlights an ongoing challenge: as these models become more capable of interacting with the real digital world, the attack surface for malicious actors expands proportionally.
