Vals CyberBench B
AI benchmark · Agentic
The same benchmark as CyberBench, listed under another name.
What is Vals CyberBench?
CyberBench measures whether AI agents can craft proof-of-concept inputs that trigger real OSS-Fuzz memory-safety vulnerabilities and patch the source code to fix them without breaking functionality.
Security mostly
“CyberBench tests models’ ability to find and trigger vulnerabilities in real-world repos, and fix them.”
vals.ai/benchmarks/cyber ↗“Can models patch the source code to fix the vulnerability, while also not affecting functionality?”
vals.ai/benchmarks/cyber ↗
Who built Vals CyberBench
Led by Vals AI.
Tier B
Established: 15 or more models scored, or curated RL directories rate it.
How much stands behind a tier: usage, who backs it, a confirmed line from its own site or paper, and a clear read on X. High with three of the four, medium with two, low otherwise.
- 20 models scored
- in curated RL directories (2 pts)
On X · mixed, low buzz, from 4 posts about Vals CyberBench in the last month
Cited as a frontier-lab cyber benchmark labs hill-climb on, but posts question fairness of access tiers, missing scores, and benchmaxxed results.
How we got to tier B
- Adoption: 20 models scored. Tier B.
- On X: mixed, low buzz. “Cited as a frontier-lab cyber benchmark labs hill-climb on, but posts question fairness of access tiers, missing scores, and benchmaxxed results.” No move.
- In curated RL directories: 2 points, a floor of C.
- Confirmed: a line from its own site or paper places it.
A clear view on X moves a tier one step up or down; most entries don't move. How tiers work.
Where Vals CyberBench sits
Its domain on the map, live. Hover a logo for its name; click to open it. Open it on the full map
Next: browse all benchmarks. Entry from the RL Research daily scrape of public sources, 2026-11-10 snapshot.