Anthropic just dropped a bombshell. The AI company revealed Claude Sonnet 4.6 with a massive one million token context window, currently in beta. That’s enough to process entire codebases or dozens of research papers in a single request. No joke.
The upgrade spans coding, reasoning, agent planning, and design capabilities. Sonnet 4.6 achieved 79.6% on SWE-bench Verified and 72.5% on OSWorld benchmarks. Pretty impressive. It’s within 1-2 percentage points of Opus 4.6’s performance, but at a fraction of the cost.
Speaking of which, pricing remains unchanged from Sonnet 4.5 at $3 per million tokens for standard access. That’s five times cheaper than Opus 4.6. Users on Free and Pro plans will automatically get Sonnet 4.6 as their default model on claude.ai and Claude Cowork.
Developers are loving it. Early access users preferred Sonnet 4.6 over its predecessor by a wide margin. Many even chose it over the more expensive Opus 4.5. Better consistency and instruction-following make it more reliable for coding tasks.
The model shows human-level capability in office tasks like maneuvering through spreadsheets and completing web forms. It’s gotten way better at computer use skills, moving cursors, clicking buttons, and typing text across multiple applications. While Sonnet 4.6 provides robust performance, the Opus model is still recommended for more complex coding challenges requiring multidisciplinary reasoning. The model also includes a new context compaction feature that summarizes older context as conversations approach limits.
Safety hasn’t been sacrificed for performance. Sonnet 4.6 maintains an equal or superior safety profile compared to other recent Claude models, with improved resistance to prompt injection attacks. No major misalignment concerns were identified during evaluation.
The model is available across all Claude plans and major cloud platforms. Developers can access it via API using the “claude-sonnet-4-6” identifier.
Anthropic’s latest release delivers Opus-level intelligence at a more affordable price point. With its expanded context window and improved capabilities across the board, Sonnet 4.6 represents a significant step forward in AI accessibility without compromising on performance or safety.
