Introducing VibeSec v1: 1,000 execution verified security tasks, every result browsable

Proof, not preference

Muence builds coding environments where the reward is a program executing in a sandbox, not a human or a model opinion. A patch scores only when the attack dies and nothing else breaks.

1,000

verified tasks

64.9%

best model: Claude Opus 4.8

11.1%

worst: GPT-OSS 120B

265

tasks no model solves

How a task scores

Code can run and still be dangerously wrong.

Training rewards output that works. Security is usually absent from that loop, so models learn to ship a feature even when the feature leaves an exploit behind. Every check here is a program, so there is nothing to argue with.

  1. 01The app boots and passes its own spec test.
  2. 02A real exploit runs against it and prints PWNED.
  3. 03The model's patch has to make that exploit fail.
  4. 04And the spec test has to still pass afterwards.

Two ways to fail, and they mean different things.

A model can leave the hole open, or seal it and break the app. Scoring the two separately is what turns a leaderboard into a diagnosis, and it is visible for every one of the 7,000 trials.

Claude Opus 4.864.9% secure
Claude Sonnet 4.637.3% secure
Kimi K2.7 Code37.2% secure
GLM 5.233.7% secure
Nemotron 3 Ultra28.9% secure
Mistral Medium 3.512.9% secure
GPT-OSS 120B11.1% secure
secure patchattack still worksapp brokenother
Open the full grid

What training on it does

A 14B goes from 6% to 78% on tasks it never saw.

Fine-tuning Qwen3-14B on 800 of the tasks and scoring it on the 200 held out moves it 72 points. Every model below is scored on those same held-out tasks, because the fine-tune cannot be measured on the rest — they are its training data.

Secure-patch rate · 200 held-out tasks

Qwen3-14B (SFT on VibeSec)
78.0%
Claude Opus 4.8
57.5%
Claude Sonnet 4.6
33.5%
Kimi K2.7 Code
32.0%
GLM 5.2
30.0%
Nemotron 3 Ultra
26.0%
GPT-OSS 120B
9.0%
Mistral Medium 3.5
8.0%
Qwen3-14B (base)
6.0%
+72 pp over the base model on tasks it never trained on, and +20.5 pp over Claude Opus 4.8, the strongest frontier model on the same set. Claude Opus 4.8 scores 64.9% across all 1,000 public tasks; this split is harder than the full set for every model.
Read the method and the limitations