↓ Ir para o conteúdo principal
Scholion
Scholion

← todas as notas

📎 Webclip

Introducing Sonnet 4.6

Claude Sonnet 4.6 is presented as the most capable Sonnet model so far, with gains across coding, computer use, long-context reasoning, agent planning, knowledge work, and design. Anthropic also says it has a 1M-token context window in beta, and that it is now the default model for Free and Pro users on claude.ai and Claude Cowork.

Reading notes
#

  • The model is described as a full upgrade across coding, computer use, long-context reasoning, agent planning, knowledge work, and design.
  • Free and Pro users now get Sonnet 4.6 as the default model in claude.ai and Claude Cowork.
  • Pricing is unchanged from Sonnet 4.5, starting at $3/$15 per million tokens.
  • Developers who got early access preferred Sonnet 4.6 over Sonnet 4.5 by a wide margin, and often over Claude Opus 4.5 as well.
  • Anthropic says Sonnet 4.6 can reach performance that previously required an Opus-class model, including on office tasks.
  • The model shows a major improvement in computer use compared with earlier Sonnet models.
  • Safety evaluations found Sonnet 4.6 to be as safe as, or safer than, other recent Claude models.
  • Anthropic says it shows strong resistance to prompt injection compared with Sonnet 4.5 and performs similarly to Opus 4.6.
  • The 1M-token context window can hold entire codebases, long contracts, or dozens of research papers in one request.
  • Anthropic says the model reasons effectively across that context and saw this in the Vending-Bench Arena evaluation.
  • Early customers reported better frontend code, financial analysis, and more polished visual outputs with better layouts, animations, and design sensibility.
  • On the Claude Platform, Sonnet 4.6 supports adaptive thinking, extended thinking, and context compaction in beta.
  • On the API, web search and fetch now automatically write and execute code to filter and process search results.
  • Code execution, memory, programmatic tool calling, tool search, and tool use examples are generally available.
  • Anthropic recommends testing across different thinking effort settings when migrating from Sonnet 4.5.
  • Opus 4.6 is still described as the strongest choice for the deepest reasoning tasks.
  • Claude in Excel now supports MCP connectors, letting Claude work with external tools and data sources from inside Excel.
  • Sonnet 4.6 is available on Claude plans, Claude Cowork, Claude Code, the API, and major cloud platforms.