Guide

Claude Sonnet 4.6

Claude Sonnet 4.6 is the sweet spot of the Claude family - strong reasoning at a speed and price that fit all-day use. For most people it is the best default for real work on Magnus.

← All guides

At a glance

Launched February 17, 2026 by Anthropic, twelve days after Opus 4.6. Sonnet 4.6 brought a major step up in coding consistency and instruction following - in Anthropic's preference testing, developers chose it over its predecessor by a wide margin - plus markedly better computer-use skills.

On the All-inclusive plan a Sonnet 4.6 reply costs 3 credits. On the API it is priced at $3 per million input tokens and $15 per million output tokens - a fifth of Opus-class pricing.

Benchmarks

  • Anthropic's headline for 4.6 was consistency, instruction following, and agentic computer use - the areas that matter most for tool-driven assistants like Magnus.
  • Preferred over Sonnet 4.5 by a wide margin in developer preference testing.

Best at on Magnus

Daily driving: chat, documents, goals, charts, and the full integration toolset (Gmail, Calendar, Slack actions, web search, PDFs). If you run one model for everything, run this one.

  • Long integration-heavy asks ("triage my inbox, then draft replies") are exactly what Sonnet 4.6's instruction-following gains were built for.