Claude Sonnet 5 launches at $2M intro pricing, default for 200M+ users; California state deal

Anthropic launched Claude Sonnet 5 on June 30, 2026, setting it as the default model for all Free and Pro Claude.ai users, with API availability across the Anthropic Platform, AWS Bedrock, Google Vertex AI, and other integrations. The model carries an introductory price of $2 per million input tokens and $10 per million output tokens through August 31, 2026, stepping to standard pricing of $3/$15 thereafter—matching Sonnet 4.6's per-token rate. Anthropic describes Sonnet 5 as its most agentic Sonnet to date, approaching Opus 4.8 performance at roughly a third of Opus's cost, with 1M token context window and new adaptive thinking capabilities.

Claude Sonnet 5 targets enterprise and developer workflows with strong coding, tool use, and multi-step agentic capabilities. Independent benchmarks show Sonnet 5 at 63.2% on agentic coding (vs. Sonnet 4.6's 58.1%), with claims of performance close to Opus 4.8 (57.4%) on reasoning-heavy tests. A critical implementation detail: Sonnet 5 ships with a new tokenizer that produces approximately 30% more tokens for the same input text, meaning even at identical per-token pricing, per-request costs can rise materially versus Sonnet 4.6.

Concurrent with the API launch, California Governor Gavin Newsom signed a first-of-its-kind state deal with Anthropic, granting all California state agencies, cities, and counties access to Claude at 50% discount beginning July 1. The state has already piloted Claude deployments (Poppy, an AI assistant for state workers; Engaged, a deliberative democracy platform; DMV and Medicaid case worker support; CalOES cybersecurity scanning). The deal is politically notable given simultaneous federal designation of Anthropic as a supply-chain risk in certain contexts.