Cognition has made Claude Mythos 5.2 available in Devin Security’s managed scans for eligible enterprise customers, the company said on X on 10 November and in a blog post. On Cognition’s internal cyber benchmark, Mythos 5.2 scored 73.5, against 56.9 for Mythos 5.1 and 32.7 for Claude Opus 5.5, the post says.
That puts Mythos 5.2 at about 2.2x Opus 5.5 and 16.6 points, or 29%, above Mythos 5.1, by our arithmetic. Cognition also said it is “giving $100M+ in credits” for scans with the model, and that “Eligibility criteria and conditions apply”.
Business Pill · The Model and the Harness
An evergreen Business Engineer explainer on how models and the products built around them split value. It teaches the general idea only and does not discuss Cognition or Mythos.
The key insight: Cognition says Claude Mythos 5.2 scored 73.5 on its internal cyber benchmark, against 56.9 for Mythos 5.1 and 32.7 for Claude Opus 5.5. That is about 2.2x Opus 5.5, by our arithmetic. Cognition is offering $100M+ in credits for scans with the model to eligible enterprise customers.
The benchmark is “a diverse set of 25 repos spanning different frameworks and languages”, each pinned to a commit where publicly reported vulnerabilities are still present, “1,781 in total”, Cognition’s post says.
The score combines how many known vulnerabilities a model finds and how many of its findings are real into an F2 score on a 0–100 scale, “weighted toward recall”. Cognition says it ran each model three times on every repo and reports the mean.
Cognition’s internal vulnerability-discovery benchmark (F2 score, 0-100, mean of three runs on 25 repos with 1,781 known vulnerabilities): Claude Mythos 5.2 73.5, Mythos 5.1 56.9, Claude Opus 5.5 32.7, as reported by Cognition. About 2.2x Opus 5.5 is our arithmetic; the scores are not independently verified.
Credits and Access
Cognition’s X post puts the credits at $100M+. Its blog post repeats the hundreds-of-millions figure it gave on 26 October, when it launched the new Devin Security: “Cognition is committing hundreds of millions of dollars in Devin credits”, offering current and future enterprise customers free scans with frontier cyber models.
Devin “runs managed scans with frontier cyber models, so your security team doesn’t need separate access to the models”, the post says. Anthropic launched Claude Fable 5.2 and Mythos 5.2 on 9 November and offers Mythos 5.2 through verification programs for cybersecurity and life-sciences work, its announcement says.
Where Devin Sits Among Mythos Routes
Security vendors are adding Mythos to their own products. IBM said on 9 November it plans to put Claude Mythos into Concert Protect, with general availability planned for early 2027. Cisco is adding it to AI Defense under existing licenses without additional cost, The Register reported, as we covered in IBM Plans Claude Mythos for 2027.
Fortune reported that Cognition’s September Series E raised more than $2 billion at a $48 billion valuation, and a Robinhood release said Cognition had recently crossed $1B in annualized revenue run rate, as we reported. On those company figures, $100M+ in credits equals at least a tenth of that run rate, by our arithmetic.
The Structural Read
As we read the posts, Cognition is selling access to a gated model through its own product: Devin runs the scans, so a security team does not need its own access to Mythos. The score that supports the pitch is Cognition’s own.
Cognition, Devin Security blog post
“Devin runs managed scans with frontier cyber models, so your security team doesn’t need separate access to the models.”
Three Implications
FOR ENTERPRISE SECURITY TEAMS Cognition says eligible enterprise customers can get free scans with frontier cyber models, including Claude Mythos 5.2, and that eligibility criteria and conditions apply.
FOR ANTHROPIC Anthropic offers Mythos 5.2 through verification programs, its announcement says; Devin, IBM and Cisco are adding routes through their own products, per Cognition, IBM and The Register.
FOR BENCHMARK READERS Cognition says it ran each model three times on 25 repos with 1,781 known vulnerabilities and reports an F2 score weighted toward recall.
The framework says: “Pricing power leaves the model and lands on whoever controls the surface and its reusable workflows.”
As we read it, Devin Security is such a surface for Mythos: Cognition packages the model inside managed scans and pays for adoption with credits. The framework does not discuss Cognition, Devin Security or Mythos.
What Is Not Established
The 73.5 score comes from Cognition’s own benchmark and has not been checked independently. Cognition has not said how the credits are split, which customers qualify, or how much of the commitment is new since 26 October. Anthropic has not published a score on this benchmark.
This Week In The AI Supercycle
The Great Convergence of Capital, Power, and Geopolitics
This week’s issue: for two years the AI build-out was priced as a supply story, and last week as a credit story. This week it was priced as a permission story, and the answer arrived as a cost, not a ban.
Cognition says Claude Mythos 5.2 scored 73.5 on its internal cyber benchmark, about 2.2x Claude Opus 5.5’s 32.7 by our arithmetic, and it is offering $100M+ in credits for scans with the model to eligible enterprise customers.
A note on sourcing. Benchmark results, methods and the credit offer come from Cognition’s blog posts (26 October and 10 November 2026) and its X posts. Context on Mythos 5.2, IBM, Cisco and Cognition’s valuation comes from our earlier pieces, which cite Anthropic, IBM, The Register, Fortune and Robinhood. The 2.2x, 29% and one-tenth figures are our arithmetic.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur (for example, How to Digitize a Family Business in Europe) and HackerNoon. Find him on LinkedIn and Substack. See how we source.
