What Claude Fable 5 is and why it's such a big deal

Anthropic's Claude model family has a new king. Claude Fable 5 belongs to the so-called Mythos class — a designation for models that outperform the previous Opus flagship. It is the first Mythos-class model that Anthropic has released for regular users. The earlier Claude Mythos Preview had been accessible since April only to a narrow group of vetted safety partners under Project Glasswing.

According to the official announcement on the Anthropic blog, Fable 5 achieves top results in virtually all tested areas: software engineering, knowledge work, computer vision, scientific research, and many more. A simple rule applies: the longer and more complex the task, the greater Fable 5's lead over the competition and over older Claude models.

Comparison with competitors: In the benchmark table published by Anthropic, Fable 5 surpasses OpenAI's GPT-5.5, Google's Gemini 3.1 Pro, and DeepSeek V4-Pro across all key metrics. The lead is particularly impressive in agent-based tasks (SWE-bench, FrontierCode) and in long-term reasoning. Analytics firm Hex announced that Fable is the first model to exceed the 90% threshold on their core analytics benchmark for complex and long-running tasks.

What Fable 5 can do and where it's already excelling

Programming: From months to days

Payment giant Stripe found during early testing that Fable 5 compressed months of development into days. In a fifty-million-line Ruby codebase, it handled a code-wide migration that would have taken an entire team over two months, in a single day. On the FrontierCode evaluation, which tests models' ability to solve difficult programming tasks while adhering to production code standards, Fable 5 achieves the highest score among all frontier models — even at moderate effort levels.

Vibe-coding platform Base44 reports that Fable excels at "generating entire applications with a single prompt" and has outstanding tool-calling capabilities. GitHub confirmed that Fable 5 handles complex, long-running programming tasks with a level of autonomy and reliability that surpasses existing benchmarks.

Science: New hypotheses and accelerated drug development

In drug development, Mythos 5 — Fable 5's sibling model without safety restrictions — accelerated certain phases of drug design by approximately tenfold. Anthropic's internal protein design experts found that the model with protein tools but without human assistance achieved results comparable to or better than experienced human researchers. Out of 14 protein targets, 9 produced strong drug development candidates, which Anthropic is currently investigating further.

Fable 5 is also the first model that consistently produces novel, compelling scientific hypotheses. In blind comparisons, Anthropic scientists preferred Mythos-generated hypotheses in molecular biology in approximately 80% of cases over hypotheses from older Opus models.

Computer vision: Pokémon on the first try

Fable 5 also sets a new standard in tasks requiring visual perception. It can extract precise numbers from detailed scientific graphs and reconstruct the source code of a web application solely from screenshots. Previous Claudes needed complex assistive tools to play Pokémon FireRed — Fable 5 completed the game with only a minimal visual interface, without any maps or navigation aids.

Price, availability, and the catch called safety

Fable 5 and Mythos 5 cost $10 per million input tokens and $50 per million output tokens. That's less than half the price of Mythos Preview, but still double that of Opus 4.8. To put it in perspective: with intensive use in Claude Code, the bill can climb to hundreds of dollars a month.

Until June 22, Fable 5 is included for free in Pro, Max, Team, and seat-based Enterprise plans. Starting June 23, Anthropic will remove it from these plans, and continued use will require purchasing credits. The company promises, however, that as soon as capacity allows, it will bring Fable 5 back into standard subscriptions.

The biggest catch: safety classifiers. Fable 5 contains built-in filters that automatically redirect responses involving cybersecurity, biology, chemistry, and model distillation to the weaker Opus 4.8. Anthropic states that filters trigger on average in fewer than 5% of sessions. It acknowledges, however, that the safety measures are deliberately set conservatively and occasionally catch harmless requests as well.

What this means for Czechia and Europe

For Czech developers, companies, and students, Fable 5 is available immediately — via the Claude API and consumption-based Enterprise plans. Full subscriptions (Pro at $20, Max at $100/$200 per month) are available worldwide, including the Czech Republic. Claude models handle the Czech language decently, though they are primarily trained on English.

The launch of Fable 5 comes as Anthropic is preparing to go public — concurrently with OpenAI and SpaceX. The company recently closed a $65 billion Series H funding round at a valuation of $965 billion, making it the world's most valuable AI startup.

At the same time, Anthropic is pushing for a global "brake" on AI development. The company warns that systems are evolving so quickly they could soon reach self-improvement capabilities. That's precisely why the first public Mythos-class model comes with strict safety guardrails — even at the cost of a worse user experience.