Anthropic on June 9 launched Claude Fable 5, its first Mythos-class model for general use, while also introducing Claude Mythos 5, a version with some safeguards removed for a limited group of cybersecurity and life sciences researchers.
“Fable 5’s capabilities exceed those of any model we’ve ever made generally available,” Anthropic said in a blog post. The company said the model delivers state-of-the-art performance across software engineering, knowledge work, vision, scientific research, and other benchmarks.
To address potential misuse, Anthropic said Fable 5 includes safeguards that redirect some cybersecurity, biology, chemistry, and model-distillation-related requests to Claude Opus 4.8 instead.
“Without safeguards, Fable 5’s capabilities in areas like cybersecurity could be misused to cause serious damage,” the company said in a statement.
Anthropic said Mythos 5 is the same underlying model as Fable 5, but with some safeguards removed. The company said Claude Mythos 5 will initially be rolled out through Project Glasswing, its initiative with the US government, replacing Claude Mythos Preview for participating organisations. Access will initially be limited to a small group of cyber defenders and infrastructure providers.
The company said Mythos 5 has “the strongest cybersecurity capabilities of any model in the world” and plans to expand access through a broader trusted access program.
On coding benchmarks, Claude Mythos 5/Fable 5 scored 80.3% on SWE-Bench Pro, 29.3% on FrontierCode (Diamond), and 88.0% on Terminal-Bench 2.1, outperforming Claude Opus 4.8, GPT 5.5, and Gemini 3.1 Pro across all three evaluations.
Anthropic priced both Fable 5 and Mythos 5 at $10 per million input tokens and $50 per million output tokens, less than half the cost of Claude Mythos Preview.
The company said Claude Fable 5 is immediately available through the Claude API and consumption-based Enterprise plans. The company is also making the model available at no additional cost to Pro, Max, Team, and seat-based Enterprise subscribers until June 22.
Starting June 23, access on those plans will require usage credits, although Anthropic said it may extend the free access period if capacity permits. The company added that it plans to eventually restore Fable 5 as a standard feature across subscription plans once sufficient computing capacity becomes available.
“We expect demand for Fable 5 to be very high, and difficult to predict,” Anthropic said.
The company highlighted several benchmark results and customer tests.
Stripe reported that Fable 5 completed a codebase-wide migration in a 50-million-line Ruby codebase in a day, a task that Anthropic said “would otherwise have taken a whole team over two months by hand.”
In knowledge work, Anthropic said Fable 5 achieved the highest score on Hebbia’s Finance Benchmark for senior-level reasoning. The company also said the model can perform vision-based tasks such as extracting data from scientific figures and rebuilding web applications from screenshots.
Anthropic reported advances in scientific research using Mythos 5. Internal protein design teams found the model accelerated parts of the drug design process by around ten times. The company said Mythos 5 independently completed tasks, including selecting binding sites, running protein-design tools, and recovering from failures.
The company also said Mythos 5 generated molecular biology hypotheses that its scientists preferred over those produced by earlier Opus-class models about 80% of the time. Some of these hypotheses have moved to experimental evaluation.
To support monitoring and abuse prevention, Anthropic announced a new policy requiring 30-day retention of business customer data for Mythos-class models and future models of similar capability. The company said the data would not be used to train future Claude models and would be deleted after 30 days in most cases.
ALSO READ: Claude Mythos Has Changed the Vulnerability Curve. Can Defenders Keep Up?
