Anthropic launches Claude Sonnet 5.5, faster and cheaper than Sonnet 5, now powering claude.ai's free tier

Anthropic released Claude Sonnet 5.5, a new mid-tier model that Anthropic says runs over 30% faster and costs up to 30% less than Sonnet 5 for most workloads, at the same list price, while outperforming it on benchmarks (no specific figures given in this coverage). Sonnet 5.5 has also replaced the previous model powering the free tier of claude.ai. Anthropic has repeated that a Haiku 5.5 model is coming in the next few weeks. This account comes from Simon Willison's write-up summarising Anthropic's announcement, not from Anthropic's own post directly.

placeholder

The extraction itself flags no primary link (no post from anthropic.com cited), so the specific "30%+ faster, 30% cheaper" and "beats it on every benchmark" figures are Anthropic's own marketing claims relayed second-hand, not independently verified numbers. "Beats it on every benchmark" is vague packaging with no benchmark names or scores attached. The framing of a free-tier "war" against a rival product (GPT-6 Luna) is commentary and positioning, not a measured fact.

1. 01Anthropic's own model card or announcement page for Sonnet 5.5, giving named benchmark scores rather than the vague 'beats it on every benchmark' claim.
2. 02Independent third-party benchmark or cost/latency tests (for example Simon Willison's pelican SVG cost comparisons already cited) confirming the 30%+ speed and cost improvements hold up outside Anthropic's own figures.
3. 03Whether Haiku 5.5 actually ships in the coming weeks as stated, and whether OpenAI responds with a free-tier or pricing change to GPT-6 Luna within the following month.