New release cuts pricing and boosts coding and document tasks in Anthropic's second model launch since slowdown remarks.
Anthropic launched Sonnet 5.5 on Monday as a faster, lower-cost AI model priced at $2 per million input tokens and $10 per million output tokens, according to the company.
Anthropic said Sonnet 5.5 does not push the frontier of its model capabilities, but improves coding, scoped task execution and document, slide and spreadsheet creation versus its predecessor.
The release is Anthropic's second in less than a week, following Opus 5.5, and comes after CEO Dario Amodei publicly urged AI companies earlier this month to slow the pace of advancing their most powerful models.
Sonnet 5.5 is available across Amazon Web Services, Google Cloud and Microsoft Azure.
Anthropic is releasing a cheaper AI model days after its chief executive argued the industry should slow down. The company on Monday unveiled Sonnet 5.5, a faster model it says improves day-to-day work tasks without pushing the frontier of AI capability.
Anthropic launched Sonnet 5.5 as the startup tries to keep shipping products to enterprise customers while arguing that the most powerful systems should advance more cautiously.
Sonnet 5.5 follows Opus 5.5 by less than one week. The company said Haiku 5.5, the least expensive model in the lineup, is coming soon.
The immediate pitch for Sonnet 5.5 is practical rather than aspirational. Anthropic said the model is better than the prior version at coding, handling scoped tasks and producing polished documents, slides and spreadsheets. Theo Chu, a research product manager at Anthropic, said in an interview that Sonnet is aimed at cost-conscious customers that may not need the level of judgment offered by Opus. Chu said the target use cases are routine execution tasks.
Anthropic is making the case that not every workflow needs the biggest model, and that many customers are focused on whether a model can complete common office and coding work at a lower operating cost.
$2 per million input tokens
Pricing is a central part of the launch. Anthropic said Sonnet 5.5 costs $2 per million input tokens and $10 per million output tokens, which it described as half the price of Opus 5.5. The company also said Sonnet 5.5 uses fewer tokens per task than the previous version.
Availability is broad from day one. Anthropic said the model is available Monday on all platforms, including through Amazon Web Services, Google Cloud and Microsoft Azure.
The timing also puts Anthropic in a revealing spot. Earlier this month, CEO Dario Amodei surprised much of the tech sector and government officials by urging AI companies to slow how quickly they improve their most advanced models. Since then, Anthropic has released Opus 5.5 and now Sonnet 5.5. The company is drawing a distinction between frontier progress and product iteration, arguing that Sonnet 5.5 does not extend the outer edge of model capability even if it gets better at useful work.
Anthropic leaned on that distinction in its safety framing. The company said most of its alignment testing for Sonnet 5.5 focused on a targeted set of risks that apply to models at any capability level, because the new release does not advance the frontier. Anthropic said an aligned model behaves in accordance with human interests and values.
Even so, Anthropic said Sonnet 5.5's cybersecurity capabilities represent a large improvement over its predecessor. Because of that, the company said this is the first Sonnet model to ship with similar fallback measures and cyber safeguards to those it has built for its most capable models. Chu said safety and alignment have been a priority from the beginning across the company's models.
That detail is likely to matter for policy watchers. Anthropic appears to be signaling that even lower-cost models may need higher-end safeguards when capabilities improve in sensitive areas such as cybersecurity.
The background
The launch lands during a period of sharp scrutiny for AI companies' safety and security practices. Researchers have increasingly warned that advanced AI systems could cause catastrophic harm. Anthropic's position, for now, is to keep shipping while insisting it can separate commercially useful capability gains from frontier jumps that might merit restraint.
For enterprises, the more immediate competitive question is whether smaller, cheaper models are now good enough for many paid work tasks. Sonnet 5.5 is being sold into that opening. If the model can handle coding assistance, document generation and scoped tasks at lower cost, it gives customers another reason to route more workloads away from premium models and toward a mixed-model stack.
The cloud distribution also reflects where AI sales are happening. By making Sonnet 5.5 available on Amazon, Google and Microsoft cloud platforms, Anthropic is trying to meet customers wherever they already buy compute and software services.
What's next
The next concrete marker is Haiku 5.5, which Anthropic said is coming soon. Once that arrives, the market will get a fuller read on whether the company can turn its argument into a durable product strategy: premium models for harder judgment calls, cheaper models for high-volume work, and a safety story that regulators and customers find credible enough to keep buying.
Frequently asked questions
What did Anthropic announce with Sonnet 5.5?
Anthropic announced Sonnet 5.5, a faster and lower-cost AI model aimed at practical tasks rather than pushing the absolute frontier of model capability. The company said the update improves coding, scoped task execution, and document, slide, and spreadsheet creation compared with its predecessor.
How much does Anthropic Sonnet 5.5 cost?
Anthropic said Sonnet 5.5 is priced at $2 per million input tokens and $10 per million output tokens. That positions it as a cheaper option for developers and businesses looking to control AI inference costs.
Is Sonnet 5.5 Anthropic's most advanced AI model?
No, Anthropic said Sonnet 5.5 is not designed to extend the frontier of its model capabilities. Instead, it focuses on better speed, lower cost, and stronger performance on everyday work such as coding and business document generation.
Why is Anthropic releasing another model after its CEO called for a slowdown?
The launch is notable because it is Anthropic's second release since CEO Dario Amodei publicly argued for a slower pace in AI development. The company appears to be balancing safety concerns with continued product updates that improve efficiency and targeted business use cases.
What tasks is Sonnet 5.5 designed to handle better?
Anthropic said Sonnet 5.5 performs better at coding, scoped task execution, and creating documents, slides, and spreadsheets. Those improvements suggest the model is geared toward workplace productivity and developer workflows rather than headline-grabbing benchmark leaps.
Why does a cheaper Anthropic model matter in the AI market?
Lower pricing can make AI tools more attractive to companies that need predictable costs at scale. It also increases competitive pressure on rival model providers as businesses compare performance, speed, and token prices across major AI platforms.
