MiniMax Unveils M3.1 Flash Model with Improved Benchmark Scores and Low API Pricing
Current overview (not historical)
MiniMax has released the M3.1 Flash inference model, reporting a score of 42 on the Artificial Analysis Intelligence Index—an increase of 13 points over its predecessor, M3. The company announced API rates of $0.15 per million input tokens and $0.60 per million output tokens, and plans to open‑source the model on October 23. With the new pricing, the cost per task is expected to drop to roughly $0.22. The announcement highlights MiniMax’s aggressive stance in China’s AI pricing competition, suggesting that the market may see even lower costs for advanced language models in the near future.
Written with AI from the reporting · updated 10/19/2026
Expand the detailed fact route and sources
Story route
A station-by-station route from the first public evidence onward; every station carries its own evidence and uncertainty.
- 1Unclassified fact stationOct 20, 00:00recorded Oct 20, 03:21single sourcerevised
MiniMax M3.1 Flash发布并公布基准测试成绩
Read original source ↗ The DecoderReplay this moment in the explorerEvidence & revisions · 1 source records
- Subject：
- minimax
- Action：
- 发布新推理模型M3.1 Flash
- Object：
- M3.1 Flash推理模型
- Occurred (fact side)：
- Oct 20, 00:00
- Published (valid side)：
- Oct 20, 02:22
- First seen (known side)：
- Oct 20, 03:21
Fact stationSingle source · one-sided claimRevisedCopy may have been revised later; time replay is bounded. Layer labels group reading only, not causation between layers.
Uncertainty marks flag where to check the sources; they are not proven causation.
Report timeline
Explore every side through dated source reports.
- Latent SpaceDeepSeek V4.1 Flash 1017 released as open-weight model with 1M token context and MIT license
Report first recorded Oct 20, 13:31
DeepSeek released V4.1 Flash 1017, a 552B total/16B active MiE model with a 1M-token context window and MIT license. It is available through DeepSeek's API and major cloud platforms. The model scores 45 on the AA Intelligence Index, tying with GLM-5.3 for second among open-weight models, and shows improved non-hallucination rates on AA-Omniscience. It ranks #1 open model in Code Arena WebDev and Agent Arena. Pricing remains at $0.30/$1.20 per million input/output tokens during peak hours.
Read original source ↗ - The DecoderRelated fact: single sourceSource report revisedMiniMax launches M3.1 Flash, slashes AI output costs to $0.22 per task
Report first recorded Oct 20, 03:21
MiniMax has released M3.1 Flash, a new reasoning model that scores 42 points on the Artificial Analysis Intelligence Index, a 13-point improvement over its predecessor M3. The model costs less than half as much per task as M3, with output priced at $0.22 per task. M3.1 Flash ties with GLM-5.3-Flash on the Intelligence Index, outperforms Gemini 3.8 Flash (41 points), and trails Deepseek V4.1 Flash (45 points). It achieves 48 percent accuracy on Humanity's Last Exam, beating both Deepseek and GLM. On the CritPt benchmark, it scores 22 percent, ahead of GLM (15 percent) and Gemini (18 percent), but behind Deepseek (24 percent). For factual accuracy on the AA-Omniscience test, M3.1 Flash answers 56 percent of questions correctly, well above GLM (28 percent) and Deepseek (43 percent). The model uses fewer output tokens than rivals, bringing the cost per task down to about $0.22. MiniMax says it runs three times faster than M3. The model has a 93 percent hallucination rate as it attempts to answer every question rather than declining when uncertain.
Read original source ↗
Event discussion
Reader opinions are hosted on GitHub Discussions; posting requires a GitHub account. Comments are not verified facts.