Trending storyDeveloping
Qwen 3.8 27B runs locally with no tools, identifies 3 of 4 games from weights alone
1 report1 sourceupdated 3 hours ago
What happened
AI digest
On November 4, Simon Willison published a notebook test of Qwen 3.8 27B running locally, checking whether the model can identify games from the neural network weights it is given, with no tools. It got Pac-Man, Pong and Breakout right, taking 46 minutes 12 seconds, 1 hour 4 minutes 54 seconds and 52 minutes 25 seconds respectively. It failed on Freeway, calling it a Pac-Man-like game.
Written by AI from the coverage · updated 2 hours ago
Coverage
Follow the reports to see the story from different sides.
Nov 4
- Simon WillisonQwen 3.8 27B on my laptop can interpret model weights, too
Simon Willison 在笔记本上实测 Qwen 3.8 27B，仅输入神经网络权重、不使用工具，成功识别四款游戏中的三款。识别 Pac-Man、Pong 和 Breakout 分别耗时 46m 12s、1h 4m 54s 和 52m 25s，Freeway 则被误判为类似 Pac-Man 的游戏。
Heat over time
Not enough continuous observations to draw a trend yet.