Claude Sonnet
Anthropic's middle-tier model. Sonnet 4.6 (2026) is the latest. Optimal balance performance/price: ~$3/M input, $15/M output. Industry favorite for coding agents (Cursor, Aider, Claude Code default).
SWE-bench Verified 70%+ — same league as Opus for many tasks. Faster latency than Opus. Default recommendation for most agentic use cases. Pricing win vs Opus: 5x cheaper for 80-90% of Opus performance. Sonnet 3.5 (June 2024) → 3.7 (Feb 2025) → 4 (May 2025) → 4.5 (Sep 2025) → 4.6 (Nov 2025) — fast iteration. Computer Use first released on Sonnet (Oct 2024). Competitor: GPT-4o (OpenAI), Gemini 2.5 Pro. Sonnet positioning: "workhorse" — most company deployments use Sonnet tier for 70%+ of traffic.