Claude Fable 5 arrives as Anthropic’s most powerful AI yet — but with controversial limits

Anthropic has launched Claude Fable 5, its most advanced AI model widely available to the public. The promise is big: stronger coding, more capable autonomous agents, better document analysis, longer tasks, and complex enterprise workflows.

But the launch also comes with controversy. Because Fable 5 is a Mythos-class model, considered more powerful and sensitive, it ships with heavy safeguards around cybersecurity, biology, chemistry, and AI development. In some cases, it does not answer directly and routes the request to Claude Opus 4.8 instead.

Quick answer: what changed?

Claude Fable 5 is Anthropic’s most capable public model so far. It is built for long-running projects, advanced coding, agents that can work for days, and complex file analysis. But for safety reasons, some sensitive requests may be blocked or answered by another model.

What is Claude Fable 5?

Claude Fable 5 is Anthropic’s new-generation model for high-level work. According to the company, it is designed for ambitious, long-running, asynchronous tasks — the kind that require planning, multiple steps, and less human supervision.

In practice, Anthropic is saying Fable 5 is not just a better chatbot. It is a model built to serve as the base for more autonomous AI agents that can plan, test their own work, delegate subtasks, and sustain projects for much longer.

The focus is coding and agents

The biggest highlight is programming.

Anthropic says Fable 5 is its most capable model for ambitious coding projects, including large migrations, complex implementations, and multi-day autonomous sessions. It can also write tests to check its own work and use vision to compare outputs against the original goal.

That matters because the AI race is moving beyond “help me with this code” into something more ambitious: “take this whole project, understand the problem, break it into steps, implement it, test it, and give me something I can review.”

That is exactly the kind of progress that could change the daily workflow of developers, startups, and software-driven companies.

AI for heavy work, not just conversation

Fable 5 also targets complex enterprise tasks. Anthropic highlights use cases such as deep analysis, research, documents, enterprise workflows, and deliverables ready for review.

Another strong point is vision. The model can understand diagrams, charts, and tables inside files and PDFs, which matters for finance, legal work, architecture, engineering, and data analysis.

In other words: Fable 5 wants to be less of a “chat assistant” and more of a “work engine.”

The safety limits

This is where the controversy starts.

Because Fable 5 belongs to the Mythos class, Anthropic says its capabilities in cybersecurity, biology, and chemistry are strong enough to create real misuse risks. For that reason, the company added stricter filters.

When a request is flagged by the safety classifiers, the answer may be automatically handled by Claude Opus 4.8. The idea is to give broad access to the new model without releasing its full power in sensitive areas.

In practice, this means users may think they are using Fable 5, but receive an answer from another model when the system detects risk.

The controversy: filters may be too conservative

The problem is that the filters appear to have gone too far in some cases.

Reports found that Fable 5 avoided answering basic biology topics, such as mitochondria, cell membranes, mRNA vaccines, and hay fever. These are common questions, often at a school level, but the system treated them as sensitive.

Anthropic argued that it chose a conservative approach to release the model earlier without increasing misuse risks. Still, this kind of false positive is frustrating, especially for students, researchers, and professionals who expected to use the model for legitimate scientific work.

Anthropic partially walked back its strategy

The company also faced criticism over a policy that could silently reduce Fable 5’s performance on certain AI development requests. After backlash, Anthropic admitted it did not get the balance right.

The biggest change: the system should now tell users when a request is refused or routed to Opus 4.8. That is essential for trust. If the AI switches models during a task, the user needs to know.

Pricing and availability

Claude Fable 5 is available through the Claude API, consumption-based Enterprise plans, and marketplaces such as AWS, Google Cloud, and Microsoft Foundry.

Anthropic lists pricing at $10 per million input tokens and $50 per million output tokens, with the existing prompt caching discount. For workloads that need U.S.-only inference, Anthropic offers that option at 1.1x pricing.

The company also made Fable 5 temporarily available at no extra cost for Pro, Max, Team, and seat-based Enterprise plans until June 22, 2026. After that, usage is expected to require credits, at least until Anthropic has enough capacity to restore the model as a standard part of subscriptions.

Why this matters

Fable 5 represents an important shift in the AI race. The discussion is no longer just about who gives better answers. The competition now involves models that can work for days, run agents, handle real code, and operate around sensitive knowledge.

But the launch also shows the dilemma of next-generation AI: the more powerful the technology gets, the harder it becomes to release everything without limits.

Anthropic is trying to balance capability and safety. But that balance is still far from perfect.

Frequently asked questions

What is Claude Fable 5?

It is Anthropic’s new advanced model, built for long-running tasks, complex coding, autonomous agents, document analysis, and heavy enterprise work.

Is it more powerful than Opus 4.8?

Yes. Anthropic positions Fable 5 as a Mythos-class model, above the Opus line in capability.

Why does it have so many restrictions?

Because Anthropic says the model has advanced capabilities in sensitive areas such as cybersecurity, biology, and chemistry, which could be misused.

What happens when a request is blocked?

In many cases, the request may be routed to Claude Opus 4.8. After criticism, Anthropic changed the experience so users should be informed when that happens.

How much does Claude Fable 5 cost?

In the API, pricing is $10 per million input tokens and $50 per million output tokens.

Is it worth using?

For simple tasks, it may be overkill. For long projects, complex coding, heavy analysis, and autonomous agents, Fable 5 could be one of the strongest options on the market.

To the DigitalRadar, Claude Fable 5 shows the next phase of AI: models that do not just answer, but work longer, act more autonomously, and raise much bigger safety debates.