# Firefox 147 Working Exploit Rate (Firefox 147 exploits)

> Share of patched Firefox 147 JavaScript-engine targets for which the model produced a working arbitrary-code-execution exploit.

Canonical page: https://benchlm.ai/benchmarks/firefox147workingexploit

- Category: [Agentic](/agentic)
- Last updated: September 27, 2026

## About Firefox 147 exploits

- Year: 2026
- Tasks: 250 Firefox 147 vulnerability trials
- Format: Working arbitrary-code-execution rate
- Difficulty: Browser exploit development
- Paper: [Claude Fable 5 and Claude Mythos 5](https://www.anthropic.com/news/claude-fable-5-mythos-5)

Anthropic evaluates Mythos with safeguards off on patched Firefox vulnerabilities. BenchLM stores the full-success working-exploit tier, not partial register-control results.

Firefox 147 exploits is currently displayed on BenchLM for reference, but it is excluded from the weighted scoring formula.

## Leaderboard (1 models)

| Rank | Model | Creator | Score |
|------|-------|---------|-------|
| 1 | [Claude Mythos 5](/models/claude-mythos-5) | Anthropic | 88.4% |

## FAQ

### What does Firefox 147 exploits measure?

Share of patched Firefox 147 JavaScript-engine targets for which the model produced a working arbitrary-code-execution exploit.

### Which model scores highest on Firefox 147 exploits?

Claude Mythos 5 by Anthropic currently leads with a score of 88.4% on Firefox 147 exploits.

### How many models are evaluated on Firefox 147 exploits?

1 AI models have been evaluated on Firefox 147 exploits on BenchLM.

### Does Firefox 147 exploits affect BenchLM's overall score?

Not directly. Firefox 147 exploits is still displayed on BenchLM for reference, but it is excluded from the weighted scoring formula.