Anthropic Releases Cheaper Sonnet 5.5, Its Second Model Since CEO Called for AI Slowdown

Anthropic released Sonnet 5.5 on Wednesday, a lower-cost model that sits below the company's most capable and expensive Claude systems. The launch is the company's second since CEO Dario Amodei published an essay and gave interviews warning that rapid AI development could pose serious societal risks if it outpaces safety work. The juxtaposition of shipping commercial products while the chief executive calls for restraint has followed every Anthropic release since those remarks.

Sonnet 5.5 targets developers running high-volume, price-sensitive tasks rather than top-end reasoning. Anthropic has not published exact per-token pricing, but the 'cheaper' positioning points at workloads where teams weigh output quality against API bills: customer support bots, document processing pipelines, and coding assistants that fire millions of inferences per day. That is the same segment where OpenAI and Google have spent the past year aggressively cutting prices, turning mid-tier models into a commodity fight.

The company's stated logic is a two-track strategy. Anthropic argues that safety-focused labs should remain at the frontier so they can shape how the technology develops, instead of ceding ground to competitors who ship with fewer public caveats. Sonnet 5.5 fits that argument: it defends Anthropic's market share on cost without moving the absolute capability frontier, which the company reserves for its flagship models.

Amodei's slowdown comments, made in a widely covered essay and follow-up interviews, put him at odds with rivals that have kept releasing models at a rapid clip. Each Anthropic launch since then invites the same scrutiny about whether the company's actions match its public caution. Anthropic has not addressed that tension directly in the Sonnet 5.5 announcement.

For developers, the practical effect is another mid-tier option in a crowded field. Cheaper models widen what teams can afford to build, particularly for applications with heavy inference volume, but switching costs matter: the deciding factor is whether Sonnet 5.5 beats similarly priced alternatives on quality. Anthropic has not yet published benchmarks comparing the model to its own flagship or to rival mid-tier offerings.

Those numbers will determine whether the cheaper label reflects a genuine bargain. The open question is how Anthropic squares its release cadence with Amodei's warnings, and whether it will publish the pricing and benchmark data developers need to make a call. Until then, Sonnet 5.5 is a positioning move as much as a product.