MiniMax Releases M3.1 Flash: Intelligence Index Jumps 13 Points, Cost Per Task Drops to $0.22
MiniMax launches a new reasoning model that significantly outperforms its predecessor while halving costs, intensifying China's AI price war.
ImportanceMaterialEvidenceE2 unreplicatedWrite-upQuick
MiniMax released its new reasoning model, M3.1 Flash, on October 19, 2026. According to data from third-party evaluator Artificial Analysis, the model scored 42 points on the Intelligence Index, a 13-point increase over its predecessor M3, while reducing the cost per task by more than half to approximately $0.22. This performance ties it with GLM-5.3-Flash, edges past Gemini 3.8 Flash, and trails Deepseek V4.1 Flash 1017, which scored 45.
The model showed significant improvement in complex reasoning tasks. On the "Humanity's Last Exam" benchmark, M3.1 Flash achieved 48 percent accuracy, surpassing both Deepseek and GLM. It also scored 22 percent on the CritPt benchmark, which tests research-level physics reasoning, ahead of GLM-5.3-Flash and Gemini 3.8 Flash. However, the model has a notable weakness: a 93 percent hallucination rate, as it attempts to answer every question rather than declining when uncertain.
MiniMax states that M3.1 Flash runs three times faster than M3 and supports up to one million tokens of context. API pricing is set at $0.15 per million input tokens and $0.60 per million output tokens. The company plans to release open weights on October 23. This launch highlights the fierce competition between Chinese AI vendors on both performance and cost, with recent head-to-head results against Deepseek suggesting the market is rapidly shifting toward lower-cost advanced models.