What happened
Anthropic released Claude Sonnet 5.5, an update to June 2026's Claude Sonnet 5. It beats its predecessor on every benchmark and tops GPT-6 Sol on knowledge work and chart understanding, scoring 56 on Artificial Analysis Intelligence Index v4.3.2.
Why it matters
Its high score suggests cheaper models are closing the gap with the biggest ones.
What to watch
The cost advantage can disappear: Anthropic warns that setting Sonnet to max reasoning effort uses more tokens than Opus and can flip Sonnet from cheaper to more expensive. Watch whether that trade-off holds when the promised Haiku 5.5 arrives within weeks.
WHO IT HITSSoftware developers and teams using Claude Code stand to get 30% faster responses and up to 30% lower costs on most tasks, though those who push the model to max reasoning effort may see the cost advantage reverse.
Summaries like this, in your inbox every morning.
Claude Sonnet 5.5 arrives just months after Claude Sonnet 5, which launched in late June 2026. Anthropic groups its models into four tiers: flagship Fable/Mythos, high-performance Opus, balanced Sonnet, and low-cost Haiku. By improving the balanced Sonnet tier on every benchmark while making it 30% faster and up to 30% cheaper for most work, Anthropic is positioning it for well-scoped everyday tasks like fixing bugs and iterating on features.
The release also highlights a trade-off inside the model itself. On the Terminal-Bench 4.0 agent benchmark, Sonnet 5.5 not only beat GPT-5.6 Sol but matched the higher-tier Claude Opus 5.5. Yet Artificial Analysis found that at maximum reasoning effort, Sonnet 5.5 uses more tokens than Opus 5.5, making it look more expensive for the performance it delivers. Anthropic's guidance is to reach for Opus on large projects, or when maxing out Sonnet would lose the balance.
For teams building with Claude, the practical question is whether Sonnet 5.5's speed and lower cost on everyday work hold up once max reasoning effort is involved, and how the picture shifts when Haiku 5.5 is expected to arrive within weeks. The pricing per token is unchanged from Sonnet 5, so the cheaper-for-most-work claim rests on the model doing more with fewer tokens in ordinary use.
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.
Omdia principal analyst Todd Thiemann surveyed 400 security leaders; the top inhibitor to AI agent identity se…
Okta launched a multivendor reference architecture, the Blueprint Alliance, for agent runtime security
Meta said on September 28 it will offer its AI models and agents to companies and developers, starting with Mu…
Chipmaker AMD agreed to acquire World Labs, the AI startup founded by industry pioneer Fei-Fei Li, for $8.2 bi…
Anthropic's Thariq Shihipar said on the Latent Space podcast that agent security may become one of the definin…
Anthropic announced Claude Sonnet 5.5 on September 28
