# Claude Sonnet [[Anthropic]]'s balanced mid-tier model, designed to optimize the tradeoff between capability and cost. Sonnet is the workhorse of the Claude lineup — recommended for most general use, including coding, analysis, and creative work — and the model that generates the majority of Anthropic's API revenue. The name follows the poetic convention: a sonnet is medium-length, tightly structured, and reliably effective. Sonnet occupies the middle of Anthropic's hierarchy between [[Claude Haiku]] (fast and cheap) and [[Claude Opus]] (maximum capability at premium cost). ## Current version: Sonnet 4.6 Released February 17, 2026. Key specifications: 1M-token context window (~750,000 words), support for both extended thinking and adaptive thinking modes, and pricing at $3/$15 per million input/output tokens. Sonnet 4.6 is notably efficient — its context window packs more usable text than Opus 4.8 due to tokenizer optimization. ## Release timeline | Version | Date | Notes | |---------|------|-------| | Claude 3 Sonnet | March 2024 | Launched with the three-tier naming system | | Claude 3.5 Sonnet | June 2024 | Outperformed Claude 3 Opus on multiple benchmarks — the first time a lower tier surpassed the one above it | | Claude 3.5 Sonnet (upgraded) | October 2024 | Significant capability bump alongside Haiku 3.5 | | Claude Sonnet 4 | May 2025 | Generation 4 launch alongside Opus 4 | | Claude Sonnet 4.6 | February 2026 | Current version | ## KB relevance The June 2024 moment when Claude 3.5 Sonnet outperformed the more expensive Claude 3 Opus on key benchmarks was significant for the AI industry: it demonstrated that capability gains were not strictly correlated with model size or compute cost, complicating the assumption that the biggest, most expensive models are always the most capable. This matters for the [[Dual-Use AI]] debate because it means military-useful AI capabilities can emerge in cheaper, more widely deployed models — not just in restricted frontier systems. Sonnet is also the model most widely deployed through Amazon Bedrock, Google Cloud, and Microsoft Foundry — the cloud partnerships that make Anthropic's technology available to enterprises and government agencies. When the [[Anthropic Government Ban]] barred federal agencies from using Anthropic products, it was primarily Sonnet deployments — not Opus — that agencies had to rip out of their workflows. ## See Also - [[Claude Opus]] — the premium high-capability tier - [[Claude Haiku]] — the fast, cost-effective tier - [[Claude Fable 5]] — the Mythos-class tier above Opus - [[Anthropic]] — the company behind the model family