Anthropic has crossed a new threshold in frontier AI. On June 9, the company released Claude Fable 5, the first publicly available model in its new Mythos class — a tier that sits above Opus in raw capability. Alongside it, Anthropic is shipping Claude Mythos 5, the same underlying model with safety guardrails lifted, available only to a small, vetted group of cyber defenders and infrastructure providers.
The launch is notable for two reasons. First, Fable 5 is the most capable model Anthropic has ever made generally available. Second, the company is using an unusually aggressive set of safeguards to make that power broadly accessible.
What “Mythos-class” actually means
Until now, Anthropic’s most powerful tier was Opus. Mythos-class models are a step above that. The first one, Claude Mythos Preview, surfaced in April but was never released to the public — its skill at finding software vulnerabilities across operating systems and browsers made a wide release too risky. Instead, Anthropic restricted it to Project Glasswing, a cybersecurity collaboration involving the U.S. government and major firms.
Fable 5 changes the equation. It is built on the same family of capabilities but wrapped in safeguards designed to make it safe for anyone to use. The name itself reflects the relationship: Fable comes from the Latin fabula, “that which is told,” a cousin of the Greek mythos. Same story, different telling — the difference between the two models is the guardrails.
The capabilities
Anthropic describes Fable 5 as state-of-the-art on nearly every benchmark it tested, with the gap over its other models widening as tasks get longer and more complex. A few highlights from early testing:
Software engineering. Stripe reported that Fable 5 compressed months of engineering work into days. In one case, it carried out a codebase-wide migration on a 50-million-line Ruby codebase in a single day — work a full team would have spent more than two months doing by hand. On Cognition’s FrontierCode evaluation, it posted the highest score among frontier models, and it did so while being more token-efficient than earlier Claude models.
Knowledge work. On Hebbia’s finance benchmark for senior-level reasoning, Fable 5 took the top score, with strong gains in document reasoning and chart and table interpretation. Trading firm IMC reported that it aced their trading-analysis evaluations almost across the board.
Vision. Fable 5 is Anthropic’s new state-of-the-art for vision tasks. It can pull precise numbers from dense scientific figures and rebuild a web app’s source code from screenshots alone. In a memorable demonstration, it beat the game Pokémon FireRed using only raw screenshots — no maps, no navigation aids — where earlier Claude models needed elaborate helper scaffolding just to play.
Memory and long-context. The model stays coherent across millions of tokens and improves its own work using persistent notes. Anthropic also showcased it building a physics-derived simulation of the solar system that predicts eclipses, autonomously playing the factory-builder Factorio, and designing a 3D-printable CAD model inside an editor it had built itself.
Science. This is where the Mythos version flexes hardest. Using Mythos 5, Anthropic’s protein-design experts reported accelerating parts of the drug-design process roughly tenfold, with the model independently choosing binding sites, running design tools, and recovering from failures. In molecular biology, scientists preferred Mythos 5’s hypotheses about 80% of the time in blind comparisons — and one of its hypotheses, a novel mechanism for an E. coli protein, was later corroborated by an independent lab working on the same problem.
The safeguards are the story
Releasing a model this capable carries real risk. Without controls, Fable 5’s cybersecurity and biology capabilities could provide meaningful “uplift” to malicious actors — help they couldn’t get from a search engine. Anthropic’s answer is a system of classifiers: separate AI systems that watch for misuse and jailbreak attempts and step in before the main model responds.
When a classifier flags a request touching cybersecurity, biology and chemistry, or model distillation, the response is quietly handed off to Claude Opus 4.8 instead, and the user is told it happened. The company deliberately tuned these classifiers to be cautious, meaning they sometimes catch harmless requests — but Anthropic says the fallback triggers in fewer than 5% of sessions. For the other 95%-plus, Fable 5 performs effectively the same as the unrestricted Mythos 5.
On the security of those safeguards, Anthropic ran an external bug-bounty program that produced no universal jailbreaks across more than 1,000 hours of testing, and external red-teaming organizations also failed to find one — though the company notes the UK’s AI Safety Institute made early progress toward one in a short testing window. Anthropic’s framing is pragmatic: completely preventing universal jailbreaks may be impossible, but the goal is to make any that remain slow and costly enough to catch before they spread
Deep Dive: Model specs, refusal responses, and fallback
Model IDs and what’s supported
Use these exact strings in your model parameter:
| Model | API model ID | Where it runs |
|---|---|---|
| Claude Fable 5 | claude-fable-5 | Claude API, Claude Platform on AWS, Amazon Bedrock, Vertex AI, Microsoft Foundry |
| Claude Mythos 5 | claude-mythos-5 | Limited release via Project Glasswing only |
| Claude Opus 4.8 | claude-opus-4-8 | All surfaces — your recommended Fable 5 fallback target |
Pricing and availability
Both Fable 5 and Mythos 5 are priced at $10 per million input tokens and $50 per million output tokens — less than half the cost of Mythos Preview. Developers can call claude-fable-5 through the Claude API.
The rollout to subscribers is staged, reflecting Anthropic’s uncertainty about demand:
- From launch through June 22, Fable 5 is included on Pro, Max, Team, and seat-based Enterprise plans at no extra cost.
- On June 23, it’s removed from those plans, and continued use requires usage credits — unless capacity allows the window to be extended.
- Eventually, once capacity catches up, Anthropic aims to restore Fable 5 as a standard part of subscription plans.
On the API and consumption-based Enterprise plans, it’s fully available now. Mythos 5 remains locked to Glasswing partners, with a trusted-access program for biology researchers coming in the weeks ahead
