Pac-Bench: One-shot Pac-Man benchmark, Claude Opus 5.5 scores 99/100
What happened
Jon Clegg 搞了个吃豆人基准测试，只给模型一句“用单个 HTML 页面写个吃豆人游戏”，然后让 Opus 5.5 自动打分。Claude Opus 5.5 通过 Claude Code 跑出了 99/100 分，生成一个 10.8 KB 的页面，花了 9 分钟，成本 1.99 美元，音效几乎还原街机。Claude Fable 5.1 拿了 96...
Coverage
Follow the reports to see the story from different sides.
- Hacker News front pagePickPac-Bench: One-shot Pac-Man benchmark, Claude Opus 5.5 scores 99/100
Jon Clegg built a Pac-Man benchmark: one prompt, one HTML page, scored automatically by Opus 5.5. Claude Opus 5.5 hit 99/100 via Claude Code at $1.99, generating a 10.8 KB page in 9 minutes with near-arcade audio. Claude Fable 5.1 scored 96 but cost $5.87. Grok 4.7 and GPT-5.6-sol scored 94 and 90; the latter cost just $0.72 in under 5 minutes. Scoring covers controls, ghost behavior, stuck detection, maze layout, and sound. The post doesn't explain why some models ran Phase 2 or how much the harness affects scores. Worth noting: this measures model-plus-toolchain combos, not bare model capability.
Heat over time
Not enough continuous observations to draw a trend yet.