Anthropic Claude Opus 5 launches as a cheaper Fable 5 alternative
Anthropic says Claude Opus 5 beats Fable 5 on coding and knowledge work while cutting prices for cost-sensitive business users.
By Jordan Bell · Startups & Deals Reporter
· 3 min read
Anthropic Claude Opus 5 is the company’s newest artificial intelligence model, and Anthropic is pitching it as both stronger and cheaper at a time when business customers are watching AI bills more closely. For retail investors tracking the AI race, the launch shows how model makers are shifting from raw capability claims toward price, usefulness and regulatory safety.
Anthropic announced the model Friday and said in a blog post that Claude Opus 5 is its best-performing and most cost-effective model across several industry benchmarks. The company said Opus 5 beats Claude Fable 5, its prior general-use model, on coding tests and knowledge-work evaluations.
The price cut is the headline for enterprise buyers. Anthropic said Opus 5 costs $5 per million input tokens and $25 per million output tokens, which is half the price of Fable 5. Tokens are the small pieces of text an AI system reads or produces, so input tokens measure what a user sends to the model and output tokens measure what the model returns.
How much does Claude Opus 5 cost?
Anthropic said Claude Opus 5 will cost users $5 per million input tokens and $25 per million output tokens. The company said that makes the model half as expensive as Claude Fable 5, while still improving performance in coding and knowledge-work evaluations.
That matters because AI software can become costly when companies run large volumes of prompts through advanced models. A lower token price can reduce the cost of customer service tools, coding assistants, document analysis and other workplace AI uses, although the actual bill depends on usage and product setup.
Anthropic’s Dianne Penn, head of product management for research, told CNBC that enterprise customers are looking for value. Penn said a cheaper model is not useful if it fails to deliver a similar level of quality.
The launch comes as Anthropic and OpenAI try to serve customers that are less willing to test expensive AI systems without a clearer return on investment, CNBC reported. Anthropic also faces lower-cost competition from Microsoft, Amazon, Google and several Chinese startups, according to CNBC.
What is Anthropic saying about safety?
Anthropic said Opus 5 is not state-of-the-art for “risky, dual-use capabilities,” including cybersecurity. Dual-use capabilities are skills that can support legitimate work but may also enable harmful activity, such as finding and exploiting software weaknesses.
The company has faced scrutiny since April, when it released Claude Mythos Preview to a limited set of users. Anthropic said at the time that Mythos Preview was highly capable in cybersecurity and performed well at discovering and exploiting software vulnerabilities.
In June, Anthropic released Mythos 5 and also introduced Fable 5, which it said was safe for general use. Anthropic said Fable 5 had stronger capabilities than any model it had previously released publicly, according to a company blog post.
Days after those June releases, Anthropic briefly disabled access to Mythos 5 and Fable 5 to comply with an export control directive from the government that cited “national security authorities,” CNBC reported. The restrictions were lifted after roughly two weeks of negotiations, allowing Anthropic to restore access to both models.
Penn told CNBC that Anthropic continues to work with government agencies as part of its pre-deployment and testing process, including for Opus 5. Anthropic said Opus 5 still trails Mythos 5 in offensive cybersecurity and biology research.
This story draws on original reporting from CNBC.
