Claude Opus 5 cost comes in at half the price of Fable 5
Anthropic priced Claude Opus 5 at $5 per million input tokens and $25 per million output tokens as enterprise AI buyers focus on ROI.
By Marcus V. Thorne · Markets Editor
· 3 min read
Anthropic set the Claude Opus 5 cost at $5 per million input tokens and $25 per million output tokens, making its newest artificial intelligence model half the price of Claude Fable 5. The company said Friday that Opus 5 is its strongest and most economical model across several industry benchmarks, a positioning aimed at enterprise customers weighing AI performance against return on investment.
Anthropic said in a blog post that Claude Opus 5 beats Claude Fable 5, the previous model it released, on evaluations tied to coding and knowledge work. The company described the model as built for daily use, while also saying it is not the leading model for “risky, dual-use capabilities,” including cybersecurity.
How much does Claude Opus 5 cost?
Claude Opus 5 costs $5 per million input tokens and $25 per million output tokens, according to Anthropic. Tokens are the units an AI system processes when reading a prompt or generating a response, so input pricing applies to what users send to the model and output pricing applies to what the model produces.
The lower price comes as Anthropic and OpenAI face more scrutiny from customers over the cost of deploying advanced AI systems. CNBC reported that companies are less willing to run open-ended experiments with expensive models unless vendors can show a clearer financial case.
Dianne Penn, Anthropic’s head of product management for research, told CNBC that customers are asking for value. “If it’s a cheaper model or a cheaper offering, but it’s not accomplishing a similar level of quality, it’s actually not useful,” Penn said.
Anthropic also faces pressure from rivals offering lower-cost AI products, including Microsoft, Amazon, Google and several Chinese startups, according to CNBC. The competitive tension has pushed leading model developers to sell both capability and efficiency as customers look for tools that can be deployed widely inside businesses without creating uncontrolled costs.
Why Anthropic is separating general work from riskier capabilities
Anthropic said Opus 5 remains behind its Mythos 5 model in offensive cybersecurity and biology research. Dual-use capabilities refer to AI functions that can serve legitimate purposes but may also be applied in harmful ways, such as finding software weaknesses that could be patched by defenders or exploited by attackers.
The company has been under closer examination since April, when it released Claude Mythos Preview to a limited group of users. Anthropic said that model was highly capable in cybersecurity and performed well at identifying and exploiting software vulnerabilities.
In June, Anthropic introduced Mythos 5 and also released Fable 5, which it said was suitable for general use. The company said at the time that Fable 5 exceeded the capabilities of all models it had previously made available to the public.
Anthropic briefly cut off access to Mythos 5 and Fable 5 days after launch to comply with a government export control directive citing national security authorities, according to CNBC. The restrictions were removed after about two weeks of negotiations, allowing the company to restore access to both models.
Penn told CNBC that Anthropic continues to work with government agencies as part of its testing and pre-deployment process, including for Opus 5. The company’s message with the new release is that enterprise users can get stronger general-purpose performance at a lower price, while more sensitive capabilities remain subject to separate evaluation and controls.
This story draws on original reporting from CNBC.
