Anthropic launches opus 5.5 with lower prices, same reach
Anthropic has launched opus 5.5, its latest flagship AI model, saying it matches Claude Fable 5.1 on most tasks while costing less to run. The release comes as pricing and safety remain central battlegrounds in the AI race.
Anthropic on Tuesday introduced opus 5.5, the first model in its Claude 5.5 family and the company’s latest high-end AI release. The launch matters because it puts lower pricing, model performance and safety testing back at the center of the fast-moving AI race.
The company said the new model performs at the level of Claude Fable 5.1 on most tasks, while costing less to run than Opus 5, the model it replaces. Anthropic also said opus 5.5 is its first release since chief executive Dario Amodei called for a slowdown in frontier AI development.
What Anthropic says opus 5.5 changes
Anthropic said opus 5.5 will cost $4 per million input tokens and $20 per million output tokens. That compares with Opus 5 at $5 per million input tokens and $25 per million output tokens. The company said the new model is 40% less expensive than Opus 5 in one description of the launch, while another company summary said it is 20% less to run than Opus 5 and 60% cheaper than Fable 5.1.
Tokens are the units AI models use to measure small words or fragments of words. In practice, that pricing can matter as much as raw benchmark scores for companies building products on top of Anthropic’s models.
Anthropic said pricing has become a key competitive advantage in the AI industry. That is why the company is framing opus 5.5 not just as a stronger model, but as a cheaper one for customers who care about deployment costs.
Performance also remains the pitch. Anthropic said opus 5.5 is at parity with Fable 5.1 on most tasks, and that the gap between its top models is narrower than benchmark charts alone suggest. The company added that being publicly available and cheaper per token makes opus 5.5 the obvious choice for many Claude users.
Safety remains part of the message
Anthropic said the Claude 5.5 models have gone through extensive alignment testing, external evaluations and safeguards for high-risk areas such as cybersecurity and biology. The company also said it is working to improve rewards for AI alignment and developing new safety training scenarios.
That emphasis comes as the wider AI industry wrestles with fears that the technology could get out of control and cause widespread harm. In the Yahoo Finance report, Amodei was quoted earlier this month saying, “We must slow the pace at which we improve the capabilities of AI models,” in an essay urging the industry to ease up on development speed.
The launch also lands after recent tensions inside the safety debate. Yahoo Finance reported that Anthropic researcher Jacob Coxon posted on X earlier this month that he was resigning, and that Anthropic alignment science lead Evan Hubinger replied that he believes there is a greater than 10% chance that the technology could “kill all humans.”
Anthropic’s move now shifts the focus to how customers respond to a model that the company says is cheaper, strong on agentic coding and still wrapped in the safety language that has become part of its identity. The next test will come in real workloads, where pricing and reliability often decide the winner faster than benchmark charts do.
