Anthropic ships a faster cheaper Sonnet that becomes the default
Reviewed by Helen Jones
On 28 September 2026 Anthropic launched Claude Sonnet 5.5, the second model in the Claude 5.5 family. It is sold as a clear upgrade on Sonnet 5: more than 30 percent faster, and up to 30 percent cheaper for most work because it usually burns fewer tokens. List prices stay at $2 per million input tokens, $10 per million output tokens, and $0.20 per million for cache reads. The API ID is claude-sonnet-5-5, with a 1 million token context window and 128,000 token max output. Anthropic reports 70.6 percent on Terminal-Bench 4.0, against 10.3 percent for Sonnet 5. Claude Code v2.1.284 made it the default Sonnet on the Anthropic API. Platform docs list five breaking changes versus Sonnet 5, including how thinking and text between tool calls are returned. It is framed as the faster, lower-cost complement to Opus 5.5 for well-scoped coding, agents, documents, slides, and spreadsheets. It is live on Claude.ai and the major clouds. Haiku 5.5 is promised in the coming weeks. Knowledge cutoff in the docs is June 2026.
Until this week, a solo founder’s Claude stack was a known bargain: Sonnet 5 for everyday coding and copy drafts, Opus when judgement actually mattered. Defaults were sticky. Prompt chains, Cursor or Claude Code sessions, and n8n-style agents assumed one request shape and one cost curve. Switching models was a conscious choice, not something the vendor did for you overnight. The default swap changes that. Same sticker price can still mean a different bill if token use drops, or a different failure mode if tool-call text now lands in thinking blocks. The Terminal-Bench leap is real for scoped bug fixes and vibe-coding, but it does not make Sonnet the right judge of offer, positioning, or legal wording. Speed without a check on request shape is how a fragile AI workflow gets worse while the product looks busier.
Analysis
This is a change to act on, not a reason to start a new product. Pin claude-sonnet-5-5 on one existing revenue-adjacent workflow (support draft, sales-page first pass, or a well-scoped coding loop), compare last week’s token spend and whether tool-call output still parses, then keep Opus for anything that sets a customer-facing promise.
Source note
Pulse published by Collab365 Spaces, reviewed by Helen Jones on . Cite as "Anthropic ships a faster cheaper Sonnet that becomes the default", Collab365 Spaces. 3 sources referenced.