Meta's Muse Wonder scores 21 on ExploitGym vs Claude models' 191-300
scaling01 · x · 2026-10-13
scaling01 shared comparison charts of Meta's Muse Wonder Full vs Anthropic models on exploit benchmarks, with the caption "I wonder what happened here."
- ExploitGym (intended-vulnerability successes): Muse Wonder Full managed only 21 (using 21 verified successes, not the 92 raw flag captures), vs Claude Opus 5 at 191, Claude Mythos 5 at 247, Claude Mythos 5.1 at 264, and Claude Opus 5.5 at 300.
- ExploitBench (percentage scores): Muse Wonder Full at 58.9%, vs Claude Opus 5 at 75% (revised from 70% in its original report), Claude Mythos 5 at 78%, Mythos 5.1 at 83%, and Opus 5.5 at 91%.
- Caveats in the charts: Anthropic results are published 6-hour runs, evaluation budgets and setups differ, ExploitBench gives partial credit, and Meta's scoring aggregation isn't confirmed identical to Anthropic's.
The overall picture: a large gap between Meta's new model and the Claude lineup on these security/exploitation benchmarks.
More from Models
- Decision models: the new Hub category that outputs numbers, not paragraphs — erhannah · 2026-10-13
- Muse Wonder Video Model Tested: 'Pretty Rough' Output, Every Narration in Australian Accent — danshipper · 2026-10-13
- Top 10 AI models' share of developer tokens plunges from 55% to 14% — GaryMarcus · 2026-10-13
- DeepSeek V4.1 Flash suspected silent update shows high code-edit failure rate — teortaxesTex · 2026-10-13
- Meta's Muse Wonder cracks the classic M riddle at 84 tok/s for $0.012 — Present-Boat-2053 · 2026-10-13
- CMU's Graham Neubig impressed using Meta's Muse Wonder to audit OpenHands usability — shuyanzh36 · 2026-10-13