WHAT NEW IN X10.28 WED

AI collections from X · 22 collections · 58,924 posts

Back to No.021 MiniMax M3.1 Flash in practice: tests and comparisons with DeepSeek V4.1 FlashUpdated 7 hours ago
minimax-m3.1IMAGE

Hands-on claim: M3.1 Flash underperforms the previous model on SWE tasks

The author says after trying both, M3.1 Flash is not better than the previous model on SWE tasks despite higher reported scores, accusing MiniMax of benchmaxxing.

tphuang
@tphuangArchived 7 days agoOriginally posted 2026.10.20
Hands-on claim: M3.1 Flash underperforms the previous model on SWE tasks
tphuang @tphuangOriginally posted 2026.10.20
Open original on X

The most recent challenge was today w/ M3.1 Flash release. I know, they claim better scores than 1017, but guess what, it's not better. If you tried both out, there is no way you can think M3.1 Flash is somehow better at SWE tasks. Minimax just benchmax way too much.

x.com/tphuang/status/2112334078175580539Archived 2026.10.20 20:26
Open original on X