Anthropic Launches Fable 5: Public Mythos-Class Model

Anthropic launched Claude Fable 5 and Claude Mythos 5 today — a Mythos-class model that tops nearly every benchmark. Fable 5 is available to the public via API and Amazon Bedrock at $10/M input and $50/M output tokens, less than half the price of Mythos Preview. Mythos 5, the unrestricted version, goes to approved cyberdefense partners through Project Glasswing. Stripe reported Fable 5 compressed months of engineering into days on a 50M-line Ruby codebase.

Key Takeaways

First publicly available Mythos-class model, state-of-the-art on nearly all benchmarks
Fable 5 has safety guardrails: sensitive queries fall back to Opus 4.8, triggers in <5% of sessions
Mythos 5 unrestricted for Project Glasswing partners — “strongest cybersecurity capabilities of any model”
50M-line Ruby migration completed in one day (would take a team 2+ months manually)
Pricing: $10/M input, $50/M output — less than half of Mythos Preview

Architecture and Guardrails

Fable 5 and Mythos 5 share the same underlying model. The difference is entirely in the safety layer. When Fable 5 detects queries about cybersecurity exploitation, biological threats, chemical weapons synthesis, or model distillation attempts, it reroutes the response to Claude Opus 4.8 instead of answering directly.

According to Anthropic’s announcement, this mechanism triggers in less than 5% of sessions. The company ran an external bug bounty with over 1,000 hours of jailbreak testing and found no universal jailbreak. The guardrails are deliberately conservative — they will block some legitimate queries — but Anthropic says it’s working to reduce false positives.

The WIRED report adds that Anthropic’s head of product management, Diane Penn, acknowledged the tradeoff: “We’re trying to make improvements in a way that’s beneficial, even if we don’t have the perfect solution for every use case to start.”

Benchmark Performance

Domain	Result
Software Engineering	Highest on Cognition’s FrontierCode evaluation
Financial Analysis	Highest on Hebbia’s Finance Benchmark (senior-level reasoning)
Vision	New state-of-the-art; rebuilt web app source from screenshots
Long Context	Maintains focus across millions of tokens; persistent memory triples performance gains vs Opus 4.8
Protein Design	Mythos 5 matches skilled humans autonomously; 9 of 14 targets yielded strong drug candidates
Genomics	Outperformed a Science journal model with 100x smaller architecture

Practical Engineering Impact

The Stripe benchmark is the most concrete evidence of Fable 5’s capability in production. A codebase-wide migration across 50 million lines of Ruby — work that would take a dedicated engineering team over two months — completed in a single day. The model also shows improved token efficiency on coding tasks.

Other demonstrations include completing Pokémon FireRed using only raw screenshots (no navigation aids or game-state tools), autonomously building factories in Factorio, and creating a full browser-based CAD editor with an integrated AI copilot. These aren’t just benchmarks — they show sustained autonomous capability on complex, multi-step tasks.

Pricing and Availability

Fable 5 is available immediately through the Anthropic API and Amazon Bedrock. The pricing structure is aggressive at $10/M input and $50/M output tokens — positioning directly against GPT-5 and Gemini Ultra.

Subscription access has a phased rollout: through June 22, Fable 5 is included in Pro, Max, Team, and seat-based Enterprise plans at no extra cost. On June 23, it moves to usage credits only, with plans to restore subscription access later. The Inc. report notes that this temporary removal is unusual and reflects the company’s balancing act between demand, safety, and infrastructure capacity.

Scientific Research Breakthrough

Mythos 5’s capabilities extend well beyond software engineering. In drug design, Anthropic’s internal protein design experts found that Mythos 5, with protein design and bioinformatics tools but no human assistance, matches or beats skilled human operators. Nine of 14 protein targets yielded strong drug design candidates currently under investigation.

In molecular biology, Mythos 5 is Anthropic’s first model to consistently produce novel, compelling scientific hypotheses. In blinded head-to-head comparisons against Opus-class models, scientists preferred Mythos’s hypotheses approximately 80% of the time. One hypothesis — a novel mechanism for an E. coli protein — was independently corroborated by a separate lab working on the same problem.

In genomics, Mythos 5 conducted novel research over a week of largely autonomous work. It assembled single-cell data for millions of cells spanning 138 animal species, designed and trained a custom ML model, and outperformed a model published in the journal Science — despite being 100 times smaller. Anthropic intends to publish these results.

Competitive Context

The launch comes as Anthropic prepares for a public market listing alongside OpenAI and SpaceX’s xAI, as CNBC notes. It also follows Anthropic’s public call last week for a coordinated brake on frontier AI development, warning that models may be approaching recursive self-improvement. The simultaneous release of the most capable public model and a warning about AI development speed creates a tension that the company and the broader industry will need to navigate carefully in the months ahead.

Sources

Operational takeaways

For technical teams, the practical value of “Anthropic Launches Fable 5: Public Mythos-Class Model” is not the headline alone, but the checklist it creates. Teams should identify which systems are exposed, which controls already exist, and which assumptions need evidence before changing architecture, tooling, or incident response priorities.

A useful next step is to separate signal from noise. Confirm the affected cloud services, review identity permissions, inspect logging coverage, and document what would trigger escalation. This keeps the response grounded in measurable risk instead of reacting to market narratives or vendor claims.