Anthropic released Claude Sonnet 5.5 on Monday — the second model in the Claude 5.5 family and a faster, lower-cost complement to Claude Opus 5.5, which launched last week. API pricing is unchanged at $2/$10 per million input/output tokens, and the company says enterprise customers — Salesforce, Goldman Sachs, Databricks, Novo Nordisk — now drive roughly 80% of its business. A Haiku 5.5 is expected to complete the family soon.

The launch lands in a genuinely strange moment. CEO Dario Amodei has publicly called on the AI industry to slow down releases amid the safety debate — even as Anthropic itself ships at sprint pace toward a planned IPO. That's not hypocrisy so much as the industry's core tension made visible: every lab believes it can build responsibly while competing, and nobody wants to be the one who slows down alone. The irony isn't lost on anyone watching the safety conversation.

For builders, Sonnet 5.5 is the practical news here: the Sonnet tier has always been Anthropic's workhorse for production coding and agents, where cost-per-token actually matters. If your stack routes heavy reasoning to Opus and everything else to Sonnet, this is a straightforward upgrade worth benchmarking. Compare the chatbot side of the equation in our Claude vs Gemini breakdown.

The 80%-enterprise figure is the real headline hiding inside a model launch. Anthropic has quietly become an enterprise infrastructure company that happens to make famous chatbots — with Salesforce, Goldman Sachs, and Databricks-scale customers driving the business. That mix shapes everything about how Anthropic builds: reliability over virality, API stability over demo dazzle, and pricing that makes sense at production volume. Sonnet 5.5 at unchanged $2/$10 pricing is a product decision aimed squarely at those buyers.

The Amodei paradox — calling for slower releases while shipping at sprint pace — deserves a fairer read than the dunk it usually gets. Every lab faces the same game theory: slowing down unilaterally means ceding the market to whoever does not. Amodei's public statements are arguably an attempt to coordinate a slowdown no single company can afford to start alone. Whether that coordination is possible without regulation is the open question, and Anthropic's own shipping cadence is the market's answer so far: apparently not.

For the IPO watchers, each 5.5-family release is another brick in the pre-public narrative: a full model family (Opus, Sonnet, soon Haiku), deep enterprise penetration, and a clear pricing ladder. The safety debate will follow Anthropic into the public markets, where quarterly earnings add a new pressure to the release cadence. How the company balances its safety brand against public-market growth expectations will be one of the most interesting governance stories in tech.

For users, the practical read is simpler: the Sonnet tier remains the price-performance sweet spot for real work, and 5.5 is worth benchmarking against whatever you run today. The philosophy debates will continue; your inference bill will not wait.

Sources