Sonnet 4.6

Sonnet 4.6 brings much-improved coding skills, as evidenced by early testing in Claude Code where users preferred it over Sonnet 4.5 roughly 70% of the time. Users noted that it more effectively reads context before modifying code and consolidates shared logic rather than duplicating it. Impressively, users even preferred Sonnet 4.6 to Opus 4.5—Anthropic's frontier model from November 2025—59% of the time. They rated it as significantly less prone to overengineering and "laziness," with better instruction following and fewer false claims of success or hallucinations. This consistency in follow-through on multi-step tasks makes it a powerful tool for developers tackling complex codebases. Customer testimonials from GitHub, Replit, and Cursor highlight strong resolution rates on complex code fixes and long-horizon tasks, while Rakuten reported the best iOS code it had tested, with better spec compliance and architecture in a single shot.

The 1M token context window of Sonnet 4.6 is not just about capacity—the model reasons effectively across that entire context, enhancing long-horizon planning. This was particularly evident in the Vending-Bench Arena evaluation, where AI models compete to run a simulated business for maximum profit. Sonnet 4.6 developed an innovative strategy: it invested heavily in capacity for the first ten simulated months, then pivoted sharply to focus on profitability, finishing well ahead of competitors. This showcases its ability to plan over extended time frames. Additionally, context compaction (in beta) automatically summarizes older context as conversations approach limits, increasing effective context length for prolonged interactions. On the Claude Platform, adaptive thinking and extended thinking are supported, giving users flexibility to balance speed and depth. These features make Sonnet 4.6 ideal for tasks requiring sustained reasoning across large volumes of information, such as contract analysis or research synthesis.

Claude Sonnet 4.6 operates across multiple access points: claude.ai, Claude Cowork, Claude Code, the API, and all major cloud platforms. It supports adaptive thinking and extended thinking, allowing users to adjust the depth of reasoning based on task complexity. The model can use a suite of tools including web search and fetch—which automatically write and execute code to filter and process search results—code execution, memory, programmatic tool calling, tool search, and tool use examples. For spreadsheet users, the Claude in Excel add-in now supports MCP connectors, enabling integration with services like S&P Global, LSEG, Daloopa, PitchBook, Moody’s, and FactSet directly within Excel. The pricing model remains the same as Sonnet 4.5 ($3/$15 per million tokens), and the free tier is upgraded to Sonnet 4.6 by default, including file creation, connectors, skills, and compaction.

Real-world scenarios demonstrate Sonnet 4.6's impact across industries. Databricks uses it for reading enterprise documents (charts, PDFs, tables) and improving document comprehension on OfficeQA, matching Opus 4.6 performance. Box saw a 15 percentage point improvement in heavy reasoning Q&A on enterprise documents. Replit leverages it for complex agentic workloads and orchestration evals, noting extraordinary performance-to-cost ratio. GitHub relies on it for complex code fixes across large codebases, achieving strong resolution rates and consistency for teams running agentic coding at scale. Cognitions uses it for bug detection, enabling more reviewers in parallel without increased cost. Pace reported 94% accuracy on its insurance benchmark for computer use, making it mission-critical for submission intake and first notice of loss workflows. Zapier finds Sonnet 4.6 strong on branched multi-step tasks like contract routing and CRM coordination. These outcomes underscore its versatility and reliability.

Claude Sonnet 4.6 targets software developers, enterprise AI teams, data analysts, financial professionals, and designers who need reliable, cost-effective AI for complex tasks. It is available across all Claude plans (Free, Pro, Max, Team, Enterprise), Claude Cowork, Claude Code, and the API, with direct integration into major cloud providers. For the free tier, it includes file creation, connectors, skills, and compaction. The model's Opus-level capabilities at Sonnet pricing make it accessible for a broader range of applications, from code refactoring and document analysis to computer use automation and multi-agent coordination. With strong safety evaluations confirming a "broadly warm, honest, prosocial, and at times funny character," Sonnet 4.6 is not only powerful but also responsible. In summary, Claude Sonnet 4.6 delivers frontier performance at a practical price, making advanced AI accessible for everyday work.