Claude Fable 5 and Claude Mythos 5
Anthropic has launched Claude Fable 5, a powerful general-use AI model, and Claude Mythos 5, a version with lifted safeguards for cybersecurity experts. However, access to both models was suspended shortly after launch due to unspecified disruptions.
Intelligence analysis by Gemini 2.5 Flash
Anthropic unveiled two new state-of-the-art AI models, Fable 5 for general use with safeguards and Mythos 5 for specialized cybersecurity applications with fewer restrictions. Both models demonstrated superior capabilities in various domains, but their public access was temporarily halted shortly after release.
Imagine a super smart robot brain that can do amazing things like write computer code, solve tough puzzles, and even play video games just by looking at the screen. A company called Anthropic made two new versions of this brain, one for everyone to use safely, and another extra powerful one for special computer security experts. But a few days after they came out, the company had to pause them because of some problems, and they are working to fix it.
Analysis
Anthropic introduced Claude Fable 5 as a new 'Mythos-class' AI model designed for general use, boasting state-of-the-art performance across numerous benchmarks including software engineering, knowledge work, vision, and scientific research. The company noted Fable 5's exceptional ability to handle longer and more complex tasks, surpassing previous models.
Alongside Fable 5, Anthropic also launched Claude Mythos 5, which is based on the same core model but with lifted safeguards in certain areas. This version is intended for a select group of cyberdefenders and infrastructure providers, initially deployed through 'Project Glasswing' in collaboration with the US government to enhance cybersecurity capabilities.
Acknowledging the inherent risks with such advanced models, Anthropic implemented conservative safeguards for Fable 5 to prevent misuse, particularly in areas like cybersecurity. These safeguards, which redirect queries on sensitive topics to a less capable model (Claude Opus 4.8), were designed to be cautious, sometimes even catching harmless requests, but triggering in less than 5% of sessions on average. The company expressed its commitment to refining these safeguards and reducing false positives as more capable models are anticipated.
The potential benefits of Fable 5 and Mythos 5 were emphasized, citing their application in securing critical software within Project Glasswing and accelerating therapeutic development in life sciences research. The models were priced at $10 per million input tokens and $50 per million output tokens, making them significantly more affordable than Claude Mythos Preview.
Despite the promising capabilities, including Fable 5's ability to autonomously perform a codebase-wide migration that would typically take months for a team (as reported by Stripe), and its superior performance in complex analytical tasks like Hebbia’s Finance Benchmark and IMC's trading-analysis evaluations, access to both models was suspended on June 12, 2026, just three days after their announcement on June 9, 2026. Anthropic apologized for the disruption and stated it was working to restore access.
Key points
- Anthropic launched Claude Fable 5 for general use and Claude Mythos 5 for cybersecurity experts.
- Fable 5 and Mythos 5 are state-of-the-art across benchmarks like software engineering, knowledge work, and vision.
- Fable 5 includes conservative safeguards, redirecting sensitive queries to a less capable model.
- Mythos 5, with lifted safeguards, was initially deployed through Project Glasswing for cyberdefenders.
- Both models were suspended shortly after launch due to unspecified disruptions.
If the access issues are resolved, Claude Fable 5 and Mythos 5 could significantly accelerate advancements in software engineering, scientific research, and cybersecurity, potentially compressing months of work into days and offering more affordable access to advanced AI capabilities. Their ability to work autonomously for longer periods suggests a future where AI handles increasingly complex tasks with minimal human intervention.
The immediate suspension of access to Fable 5 and Mythos 5, even with implemented safeguards, highlights the significant challenges and potential risks associated with deploying highly capable AI models. The conservative safeguards, while necessary, also demonstrate the difficulty in balancing advanced functionality with safety, as they can sometimes hinder legitimate use and raise concerns about access control and potential misuse.
