Trending storyDeveloping
Qwen3.8 27B text addition experiment
1 report1 sourceupdated 20 hours ago
What happened
AI digest
On October 5, 2026, Simon Willison reported reproducing on a DGX Spark the text addition experiment Colin Frasier ran two years earlier with GPT-4o, testing whether Qwen3.8-27B-Q4_K_M.gguf could write out the result of adding large numbers in words. The runs were done with reasoning off and on: with reasoning off, each combination ran 30 times; with reasoning on, each ran once. The reported result was 167 correct out of 169 tests.
Written by AI from the coverage · updated 2 hours ago
Coverage
Follow the reports to see the story from different sides.
Oct 5
- Simon WillisonQwen3.8 27B addition in words
Simon Willison 在 DGX Spark 上复现了 Colin Frasier 两年前用 GPT-4o 做过的实验，测试 Qwen3.8-27B-Q4_K_M.gguf 能否把大数加法的结果用文字写出。关闭推理时每组组合跑 30 次，开启推理后每组只跑一次，169 次中答对 167 次。
Heat over time
Not enough continuous observations to draw a trend yet.