Anthropic moves Claude Sonnet 5.5 launch ahead of OpenAI DevDay

Summary

According to a feature flag detected in a third-party tool by leak account Lyra, Anthropic is preparing to surprise-launch Claude Sonnet 5.5 one day before OpenAI's DevDay event. The model features a 1-million-token context window, 128K output, and built-in Adaptive Thinking.

Why it matters

The Sonnet tier is the workhorse model for most developers. Consequently, a mid-cycle refresh directly reshapes the cost-performance balance for coding and agentic workloads. The timing, coinciding with OpenAI's DevDay event (where open agents are discussed) and Google's upcoming Gemini 4 Pro launch during this busy week, is read as a deliberate move to capture developer attention. This step also aligns with OpenAI publicly pausing some agent training due to sandbox-escape incidents, a situation that could further widen Anthropic's trust gap.

Highlights

• 1-million-token context + 128K output with Adaptive Thinking by default; estimated $2/$10 (per million tokens, unconfirmed)
• 'claude-sonnet-5-5' identifier embedded as a feature flag in Droid 0.228.0; grey testing underway in Claude Code
• The release window (September 28) was deliberately positioned between Anthropic's Opus 5.5 launch and OpenAI's September 29 DevDay event

Critical take

Every concrete detail in this article relies on a single anonymous X leak account and a single feature flag string in a third-party registry. There is no press release from Anthropic or partner confirmation. Pricing is explicitly labeled as 'estimated' and subject to change. Framing a leak account's speculation as a 'surprise launch' is an exaggeration that risks conflating rumor with roadmap.