Claude Sonnet
Also known as: Sonnet
Sonnet has historically occupied the value-maximizing slot in Anthropic's lineup: smarter than Haiku, cheaper than Opus. The tier first appeared with Claude 3 in March 2024 and has evolved rapidly. A notable inflection came in June 2024 when Claude 3.5 Sonnet outperformed the then-flagship Claude 3 Opus on several benchmarks at half the price, signaling that mid-tier models could cross capability thresholds previously associated only with the largest models.
By February 2026, Claude Sonnet 4.6 became preferred over the previous Opus generation by a majority of developers in coding evaluations, marking the first time a Sonnet model was rated above the prior Opus for real-world engineering work. Sonnet 4.6 supports a 1 million token context window at standard pricing and is the model most commonly recommended as the default for production API integrations. It handles coding, analysis, writing, computer use, and multi-turn agentic tasks well.
From a builder perspective, Sonnet 4.6 is the sensible starting point for almost any Claude integration. It handles 90%+ of typical coding tasks at a cost that makes iterative experimentation practical. Upgrade to Opus when tasks require stronger reasoning depth or you're running autonomous agents where reliability matters more than cost. Fall back to Haiku for high-volume, low-complexity routing, classification, or real-time chat workloads.