Claude's new baseline

Anthropic has now pinned its official Claude lineup to two names: Claude Opus 4.6 and Claude Sonnet 4.6. In the company’s own model docs, those are the two “new models” presented as the current generation, with shared support for a 1 million token context window and the full Claude application programming interface feature set. (platform.claude.com) That sounds cosmetic until you remember how companies buy these systems. A team wiring Claude into customer support, coding tools, or legal review software needs one stable target model name, not a moving pile of previews, betas, and half-retired versions. (platform.claude.com) Anthropic split the pair by job. Claude Opus 4.6 is the high-end model for agents and coding, while Claude Sonnet 4.6 is the faster, cheaper model Anthropic describes as the best speed-and-intelligence balance. (platform.claude.com) The dates show how fresh this baseline is. Anthropic introduced Claude Opus 4.6 on February 5, 2026, then launched Claude Sonnet 4.6 on February 17, 2026. (anthropic.com 1) (anthropic.com 2) Claude Sonnet 4.6 is already the default model for Free and Pro users in claude.ai and Claude Cowork. That means many people using Claude in the chat product are already sitting on the new baseline even if they never touched the application programming interface docs. (anthropic.com) Anthropic is also cleaning up the knobs around these models. In the 4.6 docs, the company says “adaptive thinking” is the recommended mode, while the older thinking setting and budget token control are deprecated for Opus 4.6 and Sonnet 4.6 and will be removed in a future model release. (platform.claude.com) The same pattern shows up in context length, which is the amount of text a model can keep in view at once, like the size of its worktable. Anthropic’s March 30, 2026 release notes say the older 1 million token beta for Claude Sonnet 4.5 and Claude Sonnet 4 will stop working on April 30, 2026, and users should move to Claude Sonnet 4.6 or Claude Opus 4.6 for the full 1 million token window at standard pricing. (platform.claude.com) That tells you what “baseline” means in practice. New platform features like longer 300,000-token batch outputs, dynamic web filtering, and newer agent tooling are being described in release notes and docs around the 4.6 family, not around older Claude names. (platform.claude.com 1) (platform.claude.com 2) Anthropic’s own pricing also reinforces the division of labor. Claude Sonnet 4.6 kept Sonnet 4.5 pricing at $3 per million input tokens and $15 per million output tokens, while Claude Opus 4.6 is priced at $5 per million input tokens and $25 per million output tokens. (anthropic.com 1) (anthropic.com 2) So if you read Anthropic docs over the next few months and keep seeing Opus 4.6 and Sonnet 4.6, that is not just branding. It is Anthropic telling developers which model names are active, which controls are current, and which versions are supposed to carry the platform while older Claude variants slide toward legacy and retirement. (platform.claude.com) (platform.claude.com)