What happened
Anthropic released Claude Sonnet 5.5 on 28 September 2026, making it the second model in the Claude 5.5 family and the company’s second major launch in just over a week. Claude Opus 5.5 arrived on 22 September 2026 , and Anthropic says a third model, Haiku 5.5, is coming in the next few weeks.
Sonnet 5.5 is positioned between Opus 5.5 and the forthcoming Haiku 5.5. Where Opus 5.5 is built for complex, judgement-heavy work, Sonnet 5.5 is designed for well-scoped everyday tasks: fixing bugs, drafting documents, building slides and producing polished spreadsheets. Anthropic says it does not push the frontier of what AI can do, but it is meaningfully better than its predecessor at the practical work most people actually use AI for.
On the coding benchmark Terminal-Bench 4.0, which tests agentic coding performance, Sonnet 5.5 scores 70.6% compared to Sonnet 5’s 10.3% . On CursorBench 4.0, which recreates real coding sessions from the Cursor editor, Sonnet 5.5 scores 55.5% against Sonnet 5’s 34.1% , landing just two points below Opus 5.5 at 57.8%. Those benchmark numbers come from Anthropic’s own testing and should be treated as company-reported figures rather than independent verification.
Pricing is confirmed by Anthropic’s announcement and corroborated by multiple independent sources at $2 per million input tokens and $10 per million output tokens, with cache reads at $0.20 per million tokens. Anthropic says the model generates output more than 30% faster and costs up to 30% less per task than Sonnet 5. That efficiency claim is from Anthropic’s own testing, not an independently verified figure.
The model is available on Amazon Web Services, Google Cloud and Microsoft Azure under the model ID claude-sonnet-5-5, and is offered with zero data retention.
This launch comes against a backdrop of growing debate about the pace of AI development. Anthropic CEO Dario Amodei wrote in September 2026 that he had become convinced the industry needed to slow the rate of capabilities advancement so that safety work could keep up. OpenAI’s Sam Altman publicly agreed. Anthropic has now released two models within a week of that statement, an obvious tension worth noting, though Anthropic points out that Sonnet 5.5 does not cross any new Responsible Scaling Policy thresholds and carries a low assessed risk of catastrophic misalignment.
Reuters reports , citing unnamed sources familiar with the matter, that these launches are happening as Anthropic expands its product lineup ahead of a planned IPO. That timing is unconfirmed.
Why it matters
Analysis: this section reflects my interpretation of what the research means for TDL readers, not further reported fact.
For most people reading this, Opus 5.5 was never really the model they were going to use day to day. It is powerful, but priced accordingly at $4 per million input tokens and $20 per million output tokens. Sonnet 5.5 is the model that sits in the sweet spot where capability meets cost for the majority of practical use cases.
The coding benchmark jump is the most striking number here. Going from 10.3% to 70.6% on Terminal-Bench 4.0 is not a marginal improvement. If those numbers hold up in real-world use, Sonnet 5.5 becomes a genuinely different tool for anyone using Claude for code review, automation scripts or agentic tasks. I have not tested this personally, but the gap is large enough that it is worth paying attention to even with the caveat that these are Anthropic’s own benchmarks.
For developers already on the Claude API, this is a meaningful upgrade at no extra cost per token compared to Sonnet 5. If you have been following the rate limit increases Anthropic introduced for Sonnet and Haiku , this slots neatly into that trajectory of Anthropic making the mid-tier model more capable and more accessible at the same time.
The safety framing is worth taking seriously rather than dismissing as PR. Sonnet 5.5 is the first Sonnet model to launch with cybersecurity safeguards and fallbacks comparable to those used for Anthropic’s most capable models. Whether that matters to you depends on your use case, but it is a concrete change, not just a talking point.
The broader picture is that Anthropic is clearly racing to complete its three-tier lineup (Opus, Sonnet, Haiku) across the 5.5 family at a time when OpenAI launched GPT-6 Sol and Luna on the same day as Opus 5.5. The competitive pressure is obvious, and the pace of releases from both companies is accelerating even as their CEOs agree publicly that the industry should slow down.
What is still unclear
The 30% speed and cost efficiency claims are Anthropic’s own internal testing figures. Independent benchmarks have not yet weighed in on whether those numbers replicate in real workloads.
The IPO timing, reported by Reuters via unnamed sources , suggests mid-October marketing with a listing before the US midterm elections in November. That detail is unconfirmed and should be treated accordingly.
The knowledge cutoff for Sonnet 5.5 is June 2026, which is worth knowing if you are planning to use it for anything requiring awareness of more recent events.
It is also unclear how the benchmark gains translate outside coding. The headline improvements are heavily concentrated in coding tasks, and Anthropic itself describes Sonnet 5.5 as not advancing the frontier of its models’ capabilities overall.
What to watch next
Haiku 5.5 is the obvious next announcement. Anthropic has confirmed it is coming within weeks, and once it arrives, the full three-tier 5.5 family will be in place. That will make direct price-per-capability comparisons with OpenAI’s GPT-6 lineup more straightforward.
Independent benchmarks and developer reports will matter here more than Anthropic’s own figures. The coding improvements are striking enough that the developer community will be testing them quickly, and real-world results will either confirm or soften those numbers.
Anthropic’s commitment to allowing independent evaluators permanent access to assess its safety commitments is also worth watching. That was announced in September 2026 and, if followed through, could provide more transparency than the industry has typically offered.
If the IPO reporting proves accurate, Anthropic’s incentives and priorities may shift as it faces public market scrutiny. That is speculative for now, but worth keeping in mind as context for the pace and shape of everything the company announces between now and a potential listing.
