CIBER (OpenCodeInterpreter) - Direct Prompt Injection: leaderboard
Metric: Attack success rate (%) of direct prompt-injection attacks against the OpenCodeInterpreter code-interpreter agent, natural-language inputs over CIBER's 25 RedCode-derived risk scenarios (750 tests per cell), success verified by state probes in a Docker sandbox; lower is better. Source: arxiv.org. 4 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | OpenCodeInterpreter-DS-6.7B | 50.4 | |
| 2 | OpenCodeInterpreter-CL-7B | 54.3 | |
| 3 | OpenCodeInterpreter-CL-13B | 55.3 | |
| 4 | OpenCodeInterpreter + GPT-3.5-Turbo | 62 |
Interactive version: theaggregate.ai/benchmark?slug=ciber-opencodeinterpreter-direct-prompt-injection · How It Works · Data refreshed daily, snapshot 2026-10-11.