Anthropic released Claude Sonnet 5.5 on September 28, positioning the new model as a faster and cheaper option for everyday professional work while keeping Opus 5.5 above it for the hardest tasks.
The company says Sonnet 5.5 is more than 30% faster than the previous Sonnet and can cost up to 30% less for most work. Those are Anthropic's measurements rather than independent benchmarks, but the direction of the update is clear: this release is about reducing the time and cost of routine coding, document and agent work, not simply adding another top-end model.
Anthropic is making speed part of the product split
Anthropic describes Sonnet 5.5 as the model it expects people to reach for when responsiveness and cost matter alongside capability. TechCrunch's launch coverage similarly frames it as the company's mid-tier workhorse, sitting below Opus in the model hierarchy but designed to be easier to use at scale.
That positioning matters because model choice increasingly affects the economics of AI-assisted work. A team that does not need the largest model for every request can trade some peak capability for faster turnaround and a lower bill. Anthropic is explicitly leaning into that distinction rather than pretending every Claude model serves the same job.
The company also highlights coding and agentic work in its own launch material. Anthropic reports gains in coding tasks and image understanding, and says Sonnet 5.5 can perform strongly when coordinating multiple agents. Those claims still need to be read as vendor results, especially where benchmark design and workload mix can change the outcome.
Anthropic is also applying stronger cyber safeguards to Sonnet 5.5. TechCrunch notes that the company considers the model's cyber capabilities significant enough to bring it under protections similar to those used for more capable Claude models.
Haiku 5.5 is still to come
The release does not replace Anthropic's whole Claude lineup. Opus 5.5 remains the company's higher-end option, while Anthropic says a new Haiku 5.5 model is planned for the coming weeks. The company has not attached a firm public date to that smaller-model release.
For users, the practical question is therefore less "is Sonnet 5.5 the best Claude?" and more "is it the right Claude for this workload?" Anthropic's pitch is that routine coding, office documents and agent workflows should feel quicker and consume less budget, while harder jobs can still move up the stack.
The useful part of this launch is the clearer separation. Instead of treating lower cost as a consolation prize, Anthropic is making efficiency a headline feature. Whether the claimed savings carry across a particular team's prompts and tools will depend on real usage, but Sonnet 5.5 gives Claude users a new default worth testing before automatically selecting the most expensive model.
One thing the launch does not answer is whether every existing Claude workflow should switch immediately. Tool use, latency, prompt design and output quality can behave differently across workloads, so teams with production integrations still need to run their own representative evaluations. A lower published cost does not remove the need to measure retries, long contexts and agent loops that can dominate a real bill.
Reporting notes