‹ BackHN Continuity

Thread

Sonnet 5.5

884 points · 613 comments · D2OQZG8l5BI1S06

  1. thefourthchime · · focus · HN ↗
    It does vey well at one shotting a PacMan clone, pretty much perfect. <a href="https:&#x2F;&#x2F;jonclegg.github.io&#x2F;pacman-bakeoff&#x2F;entries&#x2F;claude-sonnet-5-5.html" rel="nofollow">https:&#x2F;&#x2F;jonclegg.github.io&#x2F;pacman-bakeoff&#x2F;entries&#x2F;claude-son...

    2nd only to Opus 5.5, which is perfect. <a href="https:&#x2F;&#x2F;jonclegg.github.io&#x2F;pacman-bakeoff&#x2F;entries&#x2F;claude-opus-5-5-r2.html" rel="nofollow">https:&#x2F;&#x2F;jonclegg.github.io&#x2F;pacman-bakeoff&#x2F;entries&#x2F;claude-opu...

    Up until very recently, all models struggled with this.

    All results: <a href="https:&#x2F;&#x2F;jonclegg.github.io&#x2F;pacman-bakeoff&#x2F;" rel="nofollow">https:&#x2F;&#x2F;jonclegg.github.io&#x2F;pacman-bakeoff&#x2F;

    1. nicce · · focus · HN ↗
      I was able to get similar with Qwen 3.8 27B with one shot. I think this game is too well in the training data.
Open on Hacker News to reply ↗

Unofficial Hacker News client; not affiliated with Y Combinator.