Anthropic has released Claude Sonnet 4.6, positioning it as a substantial upgrade focused on coding, agentic workflows, and large-context reasoning. The new model introduces a 1 million token context window in beta and is now the default option for Free and Pro users. API pricing remains consistent with the previous Sonnet 4.5 tier, starting at $3 per million input tokens and $15 per million output tokens.
According to Anthropic, Claude Sonnet 4.6 delivers performance that previously required its higher-tier Opus models. Internally, the company reports that some developers preferred Sonnet 4.6 over Claude Opus 4.5, citing more precise instruction following and fewer overly complex solutions. While those claims are based on internal testing rather than independent benchmarks, the positioning suggests Anthropic is narrowing the performance gap between pricing tiers.
A major technical shift comes with the expanded 1 million token context window. In practical terms, that scale allows Claude Sonnet 4.6 to process entire codebases, long-form legal contracts, or large research collections within a single prompt. For developers and enterprise users, larger context windows can reduce fragmentation across sessions and improve long-horizon reasoning. However, real-world performance will depend on how effectively the model maintains coherence across extremely large inputs.
Coding remains central to the update. Anthropic describes improvements in multi-step task execution and agentic behavior, meaning the model is better suited for workflows that require planning and iterative actions rather than single-response prompts. This release also aligns with recent updates to Xcode, which added support for Anthropic’s tools. That integration gives Apple developers direct access to Claude within their development environment, potentially increasing adoption among iOS and macOS teams.
Computer-use capabilities are another focus. Claude Sonnet 4.6 is designed to navigate graphical interfaces by simulating clicks and text input, without relying solely on specialized APIs. Anthropic says this brings the model closer to handling spreadsheet manipulation, multi-step web forms, and other software tasks in a way that resembles human workflows. While this opens the door to broader automation, it also raises ongoing concerns about prompt injection and misuse. The company states that safety testing shows improvements over Sonnet 4.5 and comparable resilience to more advanced Opus models.
The timing of the launch follows public tensions between Anthropic and OpenAI over advertising and model positioning. In a market where model capability, pricing transparency, and developer tooling increasingly determine market share, Claude Sonnet 4.6 appears designed to compete on performance-per-dollar rather than headline-grabbing novelty.
Claude Sonnet 4.6 is available across all Claude subscription plans and via API access. Whether its expanded context window and improved coding performance materially shift developer workflows will depend less on launch claims and more on sustained, real-world testing in production environments.
