Exclusive: Cogent Security’s VR-1.1 model outscores Claude Code and Codex on vulnerability repair

What it does

Cogent Security released VR-1.1, a new reasoning model tailored to produce fixes for recently disclosed software vulnerabilities. On a public repair benchmark commonly used to evaluate this capability, VR-1.1 outperformed Anthropic’s Claude Code and OpenAI’s Codex, two established AI models known for code generation and vulnerability repair.

Why it matters

Fixing vulnerabilities quickly and correctly remains a bottleneck in security operations. Most vulnerability management has sped up detection and triage but fixing flaws requires deep reasoning over code context. VR-1.1 improving on leading models suggests AI can now close that gap more effectively. This reduces risk exposure time and the resources needed for manual patch development. Builders and security teams could automate more of their patch workflows, lowering operational costs and shrinking vulnerability windows.

Who it is for

Security teams, DevOps engineers, and software maintainers stand to gain practical benefits. Organizations with limited patching resources can accelerate fixes, especially for newly found vulnerabilities that demand quick turnarounds. AI-assisted repair can also help firms that rely on open-source software maintainers, who often struggle with patch backlogs. Additionally, providers of vulnerability management tools might integrate similar models to boost their automated remediation features.

The catch

Benchmark success does not guarantee flawless production performance. Real-world software environments are more complex, and AI-generated fixes may require human review to avoid introducing new bugs. It also depends on how well the model adapts to diverse codebases and languages. Integration into existing security workflows and validating AI repairs at scale remain operational challenges.

What to watch next

Operators should watch how Cogent Security extends VR-1.1’s capabilities across different software ecosystems and vulnerability types. How quickly this model gets embedded into popular security platforms will influence adoption speed. Competitors may push their own models to narrow or reverse this lead, intensifying the race to automate vulnerability repair. Finally, the accuracy and safety of AI-generated patches in production environments will be a crucial practical test.

AI Quick Briefs Editorial Desk