OPEN LINK ↗
// 1h agoMODEL RELEASE
MiniMax M3.1 Flash Tops Automation Benchmarks
MiniMax officially rolls out M3.1 Flash, revealing Space Bunny as an early anonymous version that processed over 200 trillion tokens. The model claims top-tier automation performance at $0.15/$0.60 per million tokens, with nearly three-times-faster inference.
// ANALYSIS
M3.1 Flash’s real pitch is economic: frontier-like agent performance at commodity pricing. The benchmark advantage is promising, but the headline scores remain vendor-reported until independently reproduced.
- –Scores 62.2 on Automation-Bench, reportedly ahead of Opus 5.
- –Ties Opus 5 on BankerToolBench and leads 256K-context LOCA-Bench testing.
- –Costs roughly one-third as much as MiniMax M3 while delivering about three-times the speed.
- –Available through the MiniMax API, M Plan, and MiniMax Code; weights are planned for later release.
- –Developers should validate reliability on their own tool workflows before routing production automation to a newly released model.
// TAGS
minimax-m3.1-flashllmagenttool-usecomputer-useautomationbenchmarkinference
DISCOVERED
1h ago
2026-10-19
PUBLISHED
1h ago
2026-10-19
RELEVANCE
9/ 10
AUTHOR
0x0SojalSec
