News in Short

• Claude Fable 5 is now Anthropic’s most capable AI model available to general users.

• Claude Mythos 5 offers the same core model with fewer restrictions for approved cybersecurity and research partners.

• The new Claude models show major gains in coding, scientific research, vision, and long-context reasoning.

• Anthropic has added new safety systems that automatically limit access to certain high-risk capabilities.

Anthropic has launched Claude Fable 5, its most powerful AI model yet, alongside Claude Mythos 5, a restricted version for trusted cybersecurity and research organizations. According to the company, the new Claude models outperform previous generations across coding, knowledge work, scientific research, vision tasks, and long-running autonomous workflows.

The launch stands out because Anthropic is introducing not only a stronger AI system but also a new layer of safeguards designed to prevent misuse. As a result, the announcement highlights how AI companies are balancing capability with safety as frontier models become more powerful.

What Is Claude Fable 5?

Claude Fable 5 is a new Mythos-class AI model that Anthropic has made available for general use. The company says it achieves state-of-the-art performance across nearly every major benchmark it tested. The model performs particularly well in software engineering, analytical reasoning, document understanding, vision tasks, and scientific research.

According to Anthropic, the new model can work autonomously on complex projects for longer periods than previous Claude systems. As tasks become larger and more complicated, the performance gap between Fable 5 and older Claude models reportedly grows.

For users, that means better support for large projects, lengthy documents, advanced research, and multi-step workflows that require less human supervision.

What Can Claude Fable 5 Actually Do?

Software engineering is one of the biggest areas of improvement.

During testing, payment company Stripe reported that Claude Fable 5 completed a large-scale code migration across a 50-million-line codebase in a single day. The company said the same work would have required a team of engineers working manually for more than two months.

Beyond coding, Fable 5 also performs strongly in knowledge-intensive tasks. On finance and analytical reasoning benchmarks, Anthropic says the model achieved leading scores for document interpretation, chart analysis, problem-solving, and complex reasoning.

Vision capabilities have also improved significantly. The model can extract information from scientific figures, analyze screenshots, and even recreate web application code from images alone. In another example, Anthropic highlighted Fable 5’s ability to complete Pokémon FireRed using only visual screenshots and no external navigation tools.

The company also says Fable 5 shows major gains in memory and long-context performance. It can stay focused across millions of tokens and use notes to improve its own outputs during long-running tasks.

Why Is Claude Mythos 5 Different?

Claude Mythos 5 uses the same underlying technology as Fable 5. However, approved users can access some capabilities without the safeguards applied to the general version.

For now, only organizations participating in Project Glasswing can access the model. The program focuses on cybersecurity and critical infrastructure protection. Anthropic says Mythos 5 currently delivers the strongest cybersecurity capabilities of any AI model it has released.

In addition, the company plans to expand access to selected life sciences researchers through a trusted access program.

For most consumers, however, Mythos 5 remains unavailable.

Why Did Anthropic Add New Safety Systems?

The answer comes down to risk.

Anthropic says Mythos-class models can assist with advanced cybersecurity activities and sophisticated scientific research. While those capabilities can help defenders and researchers, malicious actors could also misuse them.

To address those risks, Anthropic added a new set of safety classifiers to Claude Fable 5. These systems monitor requests involving cybersecurity, biology, chemistry, and AI model distillation.

When a request falls into one of those categories, the system automatically routes the response to Claude Opus 4.8 instead of Fable 5. Users receive a notification whenever this happens.

According to Anthropic, more than 95% of Fable 5 sessions never trigger this fallback mechanism. However, some harmless requests may occasionally activate the safeguards while the company continues refining them.

The company says it deliberately tuned the safeguards conservatively at launch and plans to reduce false positives over time.

How Could These New Claude Models Impact Research?

Beyond productivity and coding, Anthropic highlighted several scientific achievements tied to Mythos 5.

According to the company, researchers accelerated parts of internal drug discovery workflows by roughly ten times using the model. Scientists also used Mythos 5 to generate biological hypotheses that they considered more compelling than hypotheses generated by earlier Claude systems.

The company further claims that Mythos 5 conducted large-scale genomics research over an extended period with minimal human guidance. During one project, the model analyzed millions of cells across 138 animal species and developed a custom machine-learning approach.

Anthropic says the resulting system outperformed a recently published scientific model despite being roughly 100 times smaller.

If validated further, these results could demonstrate how advanced AI systems contribute to scientific discovery rather than simply assisting with research tasks.

What About Privacy and Data Retention?

Alongside the model launch, Anthropic announced a new data retention policy for Mythos-class systems.

The company will require 30-day retention for business customer traffic involving Mythos-class models. According to Anthropic, it will use that data only for safety-related purposes and not for training future Claude models.

The goal is to detect emerging threats, identify new jailbreak techniques, and improve safeguard performance.

Pricing and Availability

Claude Fable 5 is available today through Anthropic’s platform and API. Pricing starts at $10 per million input tokens and $50 per million output tokens.

Meanwhile, Anthropic currently limits Claude Mythos 5 to approved cybersecurity partners and selected research organizations. The company says broader trusted access programs will arrive in the future.

Demand for Fable 5 is expected to be high. Anthropic has already begun rolling out access across multiple subscription tiers while managing capacity constraints.

Conclusion

Claude Fable 5 represents Anthropic’s biggest leap in AI capability so far. From coding and document analysis to vision and scientific research, the model pushes Claude into a new category of advanced AI systems.

At the same time, the launch highlights a growing trend in which AI companies pair powerful capabilities with stricter safeguards. Looking ahead, that balance between performance and protection may become just as important as benchmark scores. For users, Claude Fable 5 offers an early look at how the next generation of AI could work in the real world.a glimpse of where next-generation AI is heading.