What happened
Anthropic released Claude Sonnet 5.5, saying it runs 30%+ faster and costs up to 30% less for most work, priced the same as Sonnet 5 but apparently beating it on every benchmark.
Why it matters
A model that is faster and cheaper at the same price point could shift how much work teams can run for the same budget, though the cost saving is described only as up to 30% for most tasks.
What to watch
The claim rests on Anthropic's own description, and the free tier already showed a limit — the "max" thinking effort burned 128,000 tokens at a cost of $1.28 before failing to produce an SVG. Watch whether Haiku 5.5, due "in the coming weeks", lands at a competitive price.
WHO IT HITSDevelopers and product teams choosing which frontier model to build on, and budget owners comparing per-task costs, will weigh the faster-and-cheaper claim against the observed thinking-effort failure.
Summaries like this, in your inbox every morning.
Anthropic's Sonnet line has been the mid-tier option between its cheaper Haiku models and its top-end Opus models, and this release pushes that middle tier closer to the top: Sonnet 5.5 appears almost as good as Opus 5.5 on some coding tasks, including various viral 3D animation tricks, according to the article.
The most notable detail is not the benchmark claim but where the model sits. Sonnet 5.5 is now the model used for the free tier on claude.ai, while OpenAI's ChatGPT free tier uses Luna 5.6 — a gap the article reads as Anthropic currently having a much more capable free offering.
That said, capability comes with observed rough edges. Sonnet 5.5 reproduced the same bug as Opus 5.5: the "max" thinking effort pelican thought for 128,000 tokens at a cost of $1.28 before running out of tokens and failing to produce an SVG. Whether the speed and price gains hold up in everyday use, rather than in Anthropic's description, is likely to decide how quickly teams move workloads onto it. The next signal is Haiku 5.5, which Anthropic says is coming "in the coming weeks" and which the article hopes is price-competitive with GPT-6 Luna.
For example, today's edition would include:
AI-summarized, only the topics you pick: one digest a day via Email, LINE, or Slack.
Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.
Dentsu Group embedded generative AI in its accounting approval process and says it reached 99％ accuracy on the…
Nvidia launched a platform Monday that it says can monitor AI agents and quarantine suspicious ones, following…
A September 2026 guide organizes AI coding tools into three types — terminal/agent tools like Claude Code, AI-…
On September 25, MAGI Games' three debating AIs split 1-1-1 on three mini-game proposals and refused to budge…
Running Claude Code in one shared working tree produced 8 accidents and near-misses between June and September…
A tax accountant firm on a ChatGPT business plan saw its owner's account banned, which cut off all staff ChatG…
