Claude Sonnet 4.6 1M window

Anthropic’s Claude Sonnet 4.6 is not a same-day May 22 launch. Anthropic introduced the model on February 17, 2026, and said at the time that Sonnet 4.6 included a one-million-token context window in beta. Anthropic’s support-site release notes, which were updated and surfaced again on May 22, list the same feature as the main capability change. Anthropic said Sonnet 4.6 became the default model for Free and Pro users in claude.ai and Claude Cowork when it launched. The company described the release as a full upgrade across coding, computer use, long-context reasoning, agent planning, knowledge work and design. ### So what actually happened on May 22? Anthropic’s support center published or refreshed a release-notes entry visible on May 22 that pointed users to Claude Sonnet 4.6 and highlighted the one-million-token beta context window. (anthropic.com) The underlying product launch, however, dates to February 17 on Anthropic’s newsroom site. The distinction matters because the May 22 item reads like a release-note update rather than a brand-new model debut. (anthropic.com) Anthropic’s public materials already described Sonnet 4.6 and its 1M-token beta months earlier. ### What does the one-million-token window cover? Anthropic said the 1M-token window is meant to let developers process very large inputs in a single request. (support.claude.com) On Anthropic’s Sonnet page, the company says the beta is aimed at long coding and analysis sessions, including extended code editing and document review. Anthropic separately said Claude Sonnet 4 can handle “entire codebases with over 75,000 lines of code or dozens of research papers in a single request” with one million tokens on the API. (support.claude.com) That description appears on a related Anthropic product page about 1M context support. ### Where can users actually get 1M tokens? (anthropic.com) Anthropic’s Sonnet 4.6 product page says the one-million-token context window is “currently available in beta on the API only.” The same page says Sonnet 4.6 is also available on Anthropic’s own platform, Amazon Bedrock, Google Cloud Vertex AI and Microsoft Foundry, but it does not say those surfaces all expose the 1M beta in the same way. (anthropic.com) Anthropic’s API help documentation says the Claude API can ingest 1M tokens when using Claude Opus 4.7, Opus 4.6 or Sonnet 4.6, while other models remain at 200K-plus tokens. On paid Claude chat plans, Anthropic says Sonnet 4.6 supports a 500K-token context window when chatting with Claude. ### What else changed with Sonnet 4.6 besides context length? (anthropic.com) Anthropic said Sonnet 4.6 improved coding, instruction following and consistency relative to Sonnet 4.5. In its February product post, the company said early-access developers preferred Sonnet 4.6 to its predecessor “by a wide margin” and often preferred it to Claude Opus 4.5. Anthropic’s support page on training-data recency says Claude Sonnet 4.6 was trained on data up to August 2025. (support.claude.com) Anthropic also published a system card for Sonnet 4.6 saying capability evaluations showed broad gains over Sonnet 4.5 and that the model was deployed under its AI Safety Level 3 standard. ### What should developers watch next? (anthropic.com) Anthropic’s current support pages show Sonnet 4.6 remains part of the company’s active long-context lineup, alongside Opus 4.6 and Opus 4.7, for 1M-token API use. Anthropic’s pricing page excerpt on the Sonnet page lists Sonnet 4.6 at $3 per million input tokens and $15 per million output tokens, with prompt caching and batch-processing discounts. (support.claude.com) Anthropic’s next concrete updates are likely to appear first in the same support-center release notes and model FAQ pages that now track context-window limits, training cutoffs and plan-specific availability. (support.claude.com) (anthropic.com)