Anthropic launches Claude Opus 5.5 at 40% lower cost. The list price only falls 20%

Anthropic released Claude Opus 5.5 on Tuesday, the first model in a new 5.5 family. The company says it performs at the level of its Claude Fable 5.1 model on most work and costs 40% less to run than Opus 5 on typical workloads, according to its announcement. Bloomberg and the Financial Times both framed the launch as coming ahead of Anthropic's planned stock market listing. Anthropic's announcement does not mention an offering.

The number behind the 40%

The 40% figure is not a 40% price cut. Anthropic's own breakdown has three parts:

• List prices are down 20%. Input and output tokens now cost $4 and $20 per million. The company says that is 20% below Opus 5, which puts the old prices at $5 and $25.
• Cache reads are down 60%, to $0.20 per million tokens, which works out to $0.50 before. Anthropic says cache reads make up most of the cost of agent and coding work, where the same large context is fed to the model over and over.
• Fewer tokens per task. The company says Opus 5.5 uses fewer tokens to finish the same job. Combined with the lower prices, that is where the 40% comes from, at default settings in Anthropic's tests.

So the saving depends on the job. We worked two examples using only the published prices:

Neither example includes cache-write charges, and neither assumes any drop in token use. The 40% figure depends on that last part, and it only shows up if the model really finishes your specific task with fewer tokens. A business running simple customer-service replies should expect something close to the 20% line. A team running coding agents all day is where the bigger number applies.

What else changed

Anthropic says Opus 5.5 generates output more than 30% faster than Opus 5. A fast mode that runs up to 2.5 times quicker costs $8 per million input tokens and $40 per million output tokens, double the standard rate. The company is raising five-hour usage limits on its Pro, Max and Team subscription plans. It says Claude Sonnet 5.5 and Claude Haiku 5.5 will follow in the coming weeks.

On capability, Anthropic says the model leads its benchmarks in agentic coding, computer use and knowledge work. It also cautions that benchmark margins have become a less reliable guide to real-world differences at this level. Because the model is comparable to the company's Claude Mythos 5.1 in biology and cybersecurity, it ships with extra safeguards, and some specialist uses need approval through verification programs.

Why the market cares

Price is where the AI model business is being fought, and cutting cost per task while holding capability is the pitch to enterprise buyers ahead of a listing. It also bears on the AI spending story running through today's other markets news: models run on data-center capacity that is being financed with the debt bond buyers are now pricing more carefully, as we covered in our story on AI bond spreads.

Sources: Anthropic; Bloomberg; Financial Times. Old prices and the worked examples are our arithmetic from Anthropic's published figures. Disclosure: The Chronicle's newsroom uses AI tools, including Anthropic's models. This is market information, not investment advice.