Anthropic today released two versions of its most powerful model yet, splitting the same underlying AI into a hardened public release and a restricted variant earmarked for vetted users. Claude Fable 5 goes to the general market. Claude Mythos 5 goes first to the U.S. government.
The distinction between the two products isn’t capability, but what Anthropic will let each one do. “The difference is not what the model can do, but what our safeguards will allow,” an Anthropic spokesperson said, as reported by Semafor.
“Fable 5, without safeguards, would be exceptionally strong at finding and exploiting software vulnerabilities, which could meaningfully lower the cost of cyberattacks.”
Fable 5 handles that risk by routing restricted queries, covering cybersecurity and biology among other topics, to Claude Opus 4.8, Anthropic’s next-most-capable model, instead of answering directly.
Dianne Penn, Anthropic’s Head of Product Management for Research and Labs, spelled out where the line sits to Reuters. “Let’s say I’m a college student asking the model like help me find cyber vulnerabilities on X package or code,” Penn said. “The model would refuse and Fable 5 will fall back to Opus 4.8 for a response.”
Anthropic said the guardrails trigger in fewer than 5% of sessions on average. The company acknowledged they will occasionally catch harmless requests and said it plans to reduce false positives as more capable models arrive. Hackers brought in to probe Fable 5’s defenses found no successful jailbreaks, according to Anthropic.
Mythos 5 carries those same model weights but with safeguards lifted in key areas. Anthropic described it as having “the strongest cybersecurity capabilities of any model in the world.” It will roll out through Project Glasswing as an upgrade to Claude Mythos Preview, which Anthropic had previously restricted to roughly 200 organizations. The U.S. government is first in line. Anthropic said it plans to broaden Mythos 5 access over time through a systematic trusted-access program.
The Glasswing program dates to April, when Anthropic announced that the earlier Mythos model had uncovered thousands of software vulnerabilities. Existing Mythos Preview users will be able to upgrade to Mythos 5.
On price, both models come in at $10 per million input tokens and $50 per million output tokens. That is less than half what Claude Mythos Preview cost. Early customers also reported lower overall cost per task, citing improved token efficiency relative to prior Claude versions.
Partner results from early testing were striking. Stripe found that Fable 5 compressed months of engineering work into days. Inside a 50-million-line Ruby codebase, the model completed a migration in a single day that Stripe estimated would have taken a full team more than two months by hand. On Hebbia’s Finance Benchmark for senior-level reasoning, Fable 5 posted the highest score of any model tested, with gains in document-based reasoning, chart and table interpretation, and problem solving. Trading firm IMC reported the model aced its trading-analysis evaluations nearly across the board, spanning factual lookup, conceptual reasoning, root-cause analysis, and expected-value analysis.
On Cognition’s FrontierCode evaluation, Fable 5 ranks highest among frontier models even at medium effort. In a memory test using the deck-building game Slay the Spire, access to persistent file-based memory improved Fable 5’s performance three times more than it did for Opus 4.8. Fable 5 also reached the game’s final act three times more often than Opus 4.8.
In a separate vision test, Fable 5 beat Pokémon FireRed using only raw game screenshots and a minimal harness, with no maps, navigation aids, or extra game-state information. Earlier Claude models required a complex helper harness to manage the same task.
Anthropic’s valuation stands at $965 billion, exceeding that of rival OpenAI. The dual launch, priced aggressively below the model it replaces, is a clear push to widen the company’s commercial footprint across both consumer and government markets.
Information for this story was found via the sources and companies mentioned. The author has no securities or affiliations related to the organizations discussed. Not a recommendation to buy or sell. Always do additional research and consult a professional before purchasing a security. The author holds no licenses.
