Just two months ago, Anthropic said its most capable AI model was too risky for broad release.
Now the company is putting a Mythos-class model in the hands of paying customers.
On Tuesday, Anthropic announced Claude Fable 5, a new AI model built on the same capability tier as Mythos, the system that drew attention across Silicon Valley and Washington earlier this year for its ability to identify software vulnerabilities. At the time, Anthropic limited access to a small group of organizations through a cybersecurity initiative called Project Glasswing, citing concerns that the technology could be misused if released without sufficient safeguards.
Claude Fable 5 marks a significant shift in that approach.
Claude Fable 5
Announcing the launch, Anthropic said Claude Fable 5 is “a Mythos-class model that we’ve made safe for general use” and that its capabilities “exceed those of any model we’ve ever made generally available.”
The company added that Fable 5 delivers state-of-the-art performance across a broad range of benchmarks, including software engineering, scientific research, vision, and knowledge work tasks.
The launch arrives at a pivotal moment for Anthropic. The startup recently confidentially filed paperwork for an initial public offering, setting the stage for what could become one of the largest AI listings in history. Investors are watching closely to see which AI companies can translate technical advances into long-term revenue growth.
Anthropic believes Claude Fable 5 can help make that case.
According to the company, the model delivers stronger performance across software engineering and knowledge work tasks than Claude Opus 4.8, another flagship model introduced last month. Anthropic said Claude Fable 5 scored more than 10% higher than Opus 4.8 on several benchmarks.
Claude Fable 5 (Image credit: Anthropic)
The Safety Challenge Behind Claude Fable 5
The biggest question surrounding the launch is not how capable Claude Fable 5 is. It is why Anthropic now feels comfortable making a Mythos-class model broadly available after previously limiting access.
The company addressed that concern directly.
“Releasing a model this capable comes with risks. Without safeguards, Fable 5’s capabilities in areas like cybersecurity could be misused to cause serious damage.”
Anthropic said it developed new safeguards that automatically route certain high-risk requests to Claude Opus 4.8, which provides a safer response. If a user asks a question involving restricted areas such as cybersecurity attacks or dangerous biological materials, Claude Fable 5 will not answer directly.
The company acknowledged that the safeguards can occasionally block harmless requests. According to Anthropic, the protections are triggered in fewer than 5% of sessions. Anthropic said it intentionally launched the model with conservative safety settings and plans to improve accuracy and reduce false positives over time.
“What we wanted to do was to be very intentional about building new types of classifiers and new types of safety guardrails in place for this launch,” Dianne Penn, Anthropic’s head of product management for research, told CNBC.
Anthropic is releasing more than one model. Alongside Claude Fable 5, the company introduced Claude Mythos 5, an updated version of Mythos that shares the same underlying architecture. Anthropic said some of the restrictions applied to Fable 5 have been lifted in Mythos 5 for approved use cases.
From Restricted Research Project to Revenue Engine
The announcement comes as competition among leading AI companies intensifies.
Anthropic disclosed in May that its annualized revenue run rate had reached $47 billion, up from roughly $10 billion a year earlier. The company recently raised funding at a reported $965 billion valuation, placing it ahead of OpenAI’s reported $852 billion valuation from March.
OpenAI is pursuing its own path to the public markets. The company disclosed this week that it had confidentially filed IPO paperwork with regulators. Elon Musk’s SpaceX, which merged with xAI earlier this year, is expected to begin trading publicly later this week in what could become another landmark event for the AI sector.
That backdrop raises the stakes for Anthropic CEO Dario Amodei. Public market investors will be looking beyond benchmark scores and technical demonstrations. They will want evidence that Anthropic can convert advanced AI systems into products customers are willing to pay a premium for.
Claude Fable 5 is positioned squarely in that category.
The model costs $10 per million input tokens and $50 per million output tokens, double the price of Claude Opus 4.8. Anthropic argues that customers are focused less on token pricing and more on the value generated by stronger performance.
Penn said pricing remains a key consideration for customers, though many are evaluating return on investment at the task level rather than comparing raw costs.
“You just get a higher ROI by having more intelligent models,” Penn said.
For Anthropic, that argument may prove just as important as the technology itself. As the company heads toward a potential IPO, investors will be watching whether its most advanced models can justify both their premium pricing and the valuation attached to the company behind them.
