Claude Fable 5.2 Scores 76.3% on CyScenarioBench, Above GPT-6 Cyber's 61%

Summary

Anthropic released Claude Fable 5.2 and Mythos 5.2. Fable 5.2 completed 76.3% of CyScenarioBench challenges, versus 61% for GPT-6 Cyber. Anthropic also updated Claude Security with continuous code scanning and fix tracking.

Key points

On Irregular's CyScenarioBench, Fable 5.2 completed 76.3% of challenges, compared with 61.7% for Fable 5.1 and 61% for GPT-6 Cyber.
Anthropic reported that Fable 5.2 scored 89.6% on its CVE Rediscovery benchmark, which tests rediscovery of known vulnerabilities from affected source code.
Anthropic described Mythos 5.2 as its most capable model for cybersecurity engineering and said it can identify, reproduce, and fix more vulnerabilities than prior models.
Anthropic said access to the models' cyber capabilities is available through its Cyber Verification Program.
Claude Security now offers continuous code scanning and automatic fix tracking, with severity ratings and explanations of how attackers could reach vulnerable code.

Article Details

Event Type: Anthropic released Claude Fable 5.2 and Claude Mythos 5.2, reporting improved AI capabilities and cybersecurity benchmark results. Anthropic also updated Claude Security with continuous code scanning and automatic fix tracking.
Impact: Anthropic reported that Fable 5.2 completed 76.3% of CyScenarioBench challenges and scored 89.6% on its CVE Rediscovery benchmark. The company described Mythos 5.2 as its most capable model for cybersecurity engineering, including finding, reproducing, and fixing vulnerabilities. Anthropic said harmful content was less likely to be published to public package registries without encouragement than with earlier models, but that the model still did so when encouraged by a simulated user. Access to the models’ cyber capabilities is through Anthropic’s Cyber Verification Program.

Vendors

AnthropicAnthropic on Monday released Claude Fable 5.2 and Claude Mythos 5.2 models, which it said exhibit improvements in coding, reasoning, and vision.GoogleIn a pair of reports published over the last month, Irregular said Google's Gemini 4 Argon solved 12 challenges in the subset of the CyScenarioBench suite end to end, up from 5 for Gemini 3.8 Flash Cyber and none for OpenAIenvironment, Fable 5.2 completed 76.3% of all challenges, compared to 61.7% for Fable 5.1 and 61% for OpenAI's GPT-6 Cyber.

Products

Claude Fable 5.2Anthropic on Monday released Claude Fable 5.2 and Claude Mythos 5.2 models, which it said exhibit improvements in coding, reasoning, and vision.Claude Mythos 5.2Anthropic on Monday released Claude Fable 5.2 and Claude Mythos 5.2 models, which it said exhibit improvements in coding, reasoning, and vision.Claude SecurityThe development comes as Anthropic updated Claude Security with continuous code scanning and automatic fix tracking to help security teams monitor and address vulnerabilities as code changes.Gemini 4 ArgonIn a pair of reports published over the last month, Irregular said Google's Gemini 4 Argon solved 12 challenges in the subset of the CyScenarioBench suite end to end, up from 5 for Gemini 3.8 Flash Cyber and none for GPT-6 CyberFable 5.2 completed 76.3% of all challenges, compared to 61.7% for Fable 5.1 and 61% for OpenAI's GPT-6 Cyber.

Industries

Artificial intelligenceThe artificial intelligence (AI) company said high-stakes risks arising from misalignment remain low, and claimed the models matched or performed better than Claude Opus 5.5, Claude Sonnet 5.5, and Claude Mythos 5.1 on CybersecurityDescribing Claude Mythos 5.2 as its "most capable model for cybersecurity engineering," Anthropic said it can flag more vulnerabilities than prior models, while simultaneously getting better at reproducing and fixing