Anthropic has launched Claude Fable 5 Mythos AI model publicly available for the first time while introducing a new set of safeguards designed to limit access to potentially sensitive capabilities.
The release marks a significant step for the AI company, which had previously restricted its most advanced model, Claude Mythos 5, to a small group of approved organizations. With Claude Fable 5, Anthropic is now opening access to the broader public while maintaining controls over areas such as cybersecurity, biology, chemistry, and model distillation.
According to Anthropic, Claude Fable 5 shares the same underlying architecture as Claude Mythos 5. However, when users submit queries that fall into high-risk categories, the system automatically redirects those requests to Claude Opus 4.8.
The company said more than 95% of user sessions are expected to operate without triggering these restrictions, allowing most users to experience the model’s full capabilities.
The launch comes amid growing competition in the artificial intelligence industry, where companies are racing to build increasingly powerful systems while facing pressure from governments and safety researchers to implement stronger safeguards.
Early testing partners reported significant productivity gains. Financial technology company Stripe said Claude Fable 5 completed a large-scale codebase migration involving approximately 50 million lines of Ruby code in a single day. According to the company, the same task would typically require more than two months of engineering work.
The company further claimed that Claude Fable 5 achieved industry-leading results on financial reasoning evaluations, although independent benchmark comparisons have yet to be widely published.
Alongside the launch, Anthropic introduced a mandatory 30-day data retention policy for all Mythos-class model traffic. The policy applies whether users access the model directly through Anthropic’s platforms or through third-party integrations.
Anthropic said retained conversations will not be used for model training and will generally be deleted after 30 days. The company acknowledged that the requirement may raise concerns among privacy-focused users but argued that temporary retention is necessary to monitor misuse and evaluate the effectiveness of safety systems.
The safeguards themselves represent one of the most closely watched aspects of the release.
Anthropic described Claude Fable 5 Mythos AI as the most capable AI model it has ever made generally available. The company highlighted strengths in software engineering, knowledge work, reasoning, and visual understanding.
Source: Anthropic
Anthropic stated that external security researchers spent more than 1,000 hours attempting to bypass the model’s protections. According to the company, researchers failed to discover a universal jailbreak capable of consistently circumventing the safeguards.
However, Anthropic noted that evaluators from the UK AI Safety Institute highlighted the ongoing challenge of securing advanced AI systems against misuse.
The company admitted that some harmless prompts may occasionally trigger the safety filter, which redirects users to Claude Opus 4.8. Anthropic described the current system as a trade-off between usability and security, adding that future updates could improve the precision of its classifiers.
Anthropic also pointed to the model’s performance in gaming and reasoning benchmarks. During internal testing, Claude Fable 5 successfully completed Pokémon FireRed using only a minimal vision-based setup, which was a challenge for previous Claude models.
Source: Youtube
Pricing for Claude Fable 5 matches that of Claude Mythos 5. Anthropic has set rates at $10 per million input tokens and $50 per million output tokens.
For users subscribed to Pro, Max, Team, and Enterprise plans, access to Claude Fable 5 will be included at no additional cost until June 22. After that date, usage credits will be required.
Meanwhile, Claude Mythos 5 remains restricted to organizations approved under Project Glasswing. Anthropic said it is working with the U.S. government on a broader trusted-access framework to expand availability to additional organizations.
The company also plans to create a separate access program for researchers requiring unrestricted capabilities for scientific work.
The launch of Claude Fable 5 Mythos AI underscores a broader trend in the AI industry. One, where companies are increasingly trying to balance capability with control. As models become more powerful, firms face growing scrutiny over how advanced systems are deployed and who should be allowed to use them.
For Anthropic, Claude Fable 5 represents both a technological milestone and a test of whether highly capable AI systems can be released to the public without compromising safety. The answer could influence how future generations of advanced AI models are developed and distributed across the industry.
Keep your eyes on Trending Eyes for real-time insights and analysis.
Read more recent news
Football World Cup 2026 Set for Historic Kickoff Between Messi and Ronaldo
Cockroach Janta Party Storms Indian Politics
Indian Railways Tenders 2026: Big Wins in IREPS for Contractors
FCI Tenders 2026: Food Procurement, Storage & Supply Chain Projects
NASA Artemis II Moon Mission Takes Off
Chinese Humanoid Robot Commercialization Takes Center Stage at CES 2026
HAL Shares After Tejas Grounding Tumble as IAF Freezes Tejas Fleet
