🔍 Read the full analysis: MiniMax M3.1 Flash Trails The Open-Weight Leaderboard. How Did It Do In Our Test? on ThorstenMeyerAI.com
Prime made for students and young adults
- Fast, free delivery for dorm and study essentials
- Prime Video and Amazon Music included
- Member-only deals
TL;DR
MiniMax released M3.1 Flash on October 19, 2026, pricing it at $0.15 per million input tokens and $0.60 per million output tokens. Artificial Analysis gave it a 43 Intelligence Index score, below several open-weight competitors, while the source’s newsroom test covered 24 articles but provides no scores in the supplied material.
MiniMax released M3.1 Flash on October 19, positioning the model as a low-cost option for agent work, but an independent index places it behind several open-weight competitors. A newsroom evaluation described by ThorstenMeyerAI.com tested the model on 24 real news articles; the supplied source material does not include its scores, so that test’s results cannot be assessed here.
MiniMax lists M3.1 Flash at $0.15 per million input tokens and $0.60 per million output tokens, half the per-token prices listed for its M3 predecessor. It also charges $0.015 per million cached input tokens. The model is available through the MiniMax API, M Plan subscription, MiniMax Code and OpenRouter; OpenRouter offered it free for its first 72 hours.
The company says the model delivers more than 200 tokens per second in everyday use, with a median above 170, and is “nearly three times faster” than M3. Those figures are MiniMax’s claims. Artificial Analysis measured 120 tokens per second and 1.08 seconds to first token in its testing. The service lists a one-million-token context window; OpenRouter permits up to 524,288 output tokens, and reasoning effort can be adjusted.
Artificial Analysis assigned M3.1 Flash an Intelligence Index score of 43. In the comparison cited by ThorstenMeyerAI.com, MiMo-V2.6-Pro scored 46, while DeepSeek V4.1 Flash 1017 and GLM-5.3 each scored 45. The index recorded DeepSeek at 211 tokens per second, compared with 120 for MiniMax. Artificial Analysis counted 180 million tokens used by M3.1 Flash to complete its index, against a median of 99 million; the source notes that higher token use can offset a low price per token.
Infographic 01 · MiniMax M3.1 Flash
A step behind on the index. Ahead of every open model in our newsroom.
An independent intelligence index puts MiniMax M3.1 Flash a step behind the open-weight leaders. Writing real news from real sources, it beat all of them.
Independent index
Artificial Analysis Intelligence Index, 10 evaluations. Bar length on a 0–50 scale.
Artificial Analysis lists M3.1 Flash as proprietary until its weights are released.
Our newsroom test
Mean score, English articles, 1–5 by GPT-6.1 Sol and Claude Opus 5.5 with the source in context. Bar length on a 0–5 scale.
In German, M3.1 Flash (3.36) also leads every open model, but GPT-6 Luna is clearly ahead (3.62).
Sources: Artificial Analysis, MiniMax-M3.1-Flash and open-weights models (20 Oct 2026); Thorsten Meyer AI newsroom benchmark, 12 sources × English + German, run 20 Oct 2026.
Powered by Thorsten Meyer AI
Price Against Measured Performance
M3.1 Flash’s appeal rests on a trade-off: low listed prices and a large context window versus performance that does not lead the comparison cited in the report. For developers running agent tasks at scale, the per-token rate is only one part of cost. Speed, how many tokens a model uses to finish a task and the quality of its output can change the overall economics.
The independent results also put MiniMax’s speed claims in perspective. Artificial Analysis measured a lower rate than the company’s stated everyday-use figure, and the model used more tokens than the index median. That does not establish how it will perform across every workload, but it gives prospective users independent measurements to weigh against vendor claims.
The newsroom test addresses a different practical question: whether the model can turn source material into accurate, useful articles without inventing facts. The source says two judges reviewed English and German versions of 12 real-news sources, scoring fidelity, fabrication, clarity, usefulness, structure and polish. Since the scores and model comparisons are missing from the supplied material, no conclusion about its newsroom performance can be drawn from this account.
Popular right now
De'Longhi Magnifica Duo Fully Automatic Espresso Machine with Milk Frother, Silver
Bose QuietComfort True Wireless Earbuds with Active Noise Cancellation, Bluetooth 5.3, USB-C Charging, and IPX4 Water Resistance
Roborock Robot Vacuum and Mop with LiDAR Navigation and Self-Empty Dock
As an affiliate, we earn on qualifying purchases.
From Space Bunny to M3.1
Before the public release, the model appeared anonymously on OpenRouter as Space Bunny Alpha. According to ThorstenMeyerAI.com, it went live on September 23, led OpenRouter’s usage ranking for 10 days, processed 56.7 trillion tokens on the platform and was pulled on October 5. MiniMax says the model processed more than 200 trillion tokens overall.
The report says that during the anonymous period, the provider could retain prompts but could not use them to train models. MiniMax has not said what happens to that retained data. The episode makes the service’s data-handling terms relevant to organizations that used the anonymous endpoint, although the supplied information does not establish how many users were affected or what data was retained.
MiniMax’s own benchmark charts show a mixed picture, according to the source’s transcription. M3.1 Flash scored 62.2 on Automation-Bench, ahead of DeepSeek V4.1 Flash 1017’s 60.5. DeepSeek scored higher on DeepSWE, 77.0 to 73.7, and Terminal-Bench, 42.3 to 38.1. These are company-reported results; MiniMax selected the rivals, and the source says they have not been independently reproduced. The weights remain unreleased, and MiniMax has promised them for October 23.
““nearly three times faster” than M3”
— MiniMax
Weights, Data and Test Scores
M3.1 Flash’s weights had not been released as of the October 20 report, and no license or official parameter count was available. A parameter count reported by another outlet was not confirmed by MiniMax and was excluded by the source. The promised October 23 release could clarify the model’s open-weight status and provide details, but that release is not confirmed in the supplied material.
The supplied report text ends before the newsroom benchmark’s results table. It describes the test setup and judges but gives no scores for M3.1 Flash or comparison models. It is also unclear whether MiniMax will explain its handling of prompts retained during the Space Bunny period. MiniMax’s benchmark results and security claims, including its reported CyberGym result, are company evaluations rather than independently reproduced findings in the source.
The October 23 Weights Release
The next stated milestone is MiniMax’s promised October 23 weights release. Its timing, license and the model details that may accompany it remain unconfirmed in the supplied material. If released, those details will help establish whether developers can download and run the model independently rather than access it only through hosted services.
Readers evaluating the model can also look for publication of the complete newsroom benchmark table and independent replications of MiniMax’s claims. For users who tried Space Bunny Alpha, clarification from MiniMax about retained prompts would address a separate outstanding question. Until those details are available, the strongest confirmed comparison remains the independent index score and its reported speed and token-use measurements.
Key Questions
What is MiniMax M3.1 Flash?
M3.1 Flash is a model released by MiniMax on October 19, 2026, marketed for fast, low-cost agent work. It is available through MiniMax services and OpenRouter.
How much does M3.1 Flash cost?
MiniMax lists prices of $0.15 per million input tokens, $0.60 per million output tokens and $0.015 per million cached input tokens. OpenRouter offered it free for the first 72 hours after launch.
How did it compare with open-weight models?
Artificial Analysis gave it an Intelligence Index score of 43, below the scores cited for MiMo-V2.6-Pro at 46 and DeepSeek V4.1 Flash 1017 and GLM-5.3 at 45. The comparison is specific to that index, not every task or use case.
What did the newsroom test find?
The report describes a test of 24 English- and German-language articles, reviewed by two judges for source fidelity and other writing qualities. The supplied material does not include the results table, so it does not establish how M3.1 Flash scored.
Are the model weights available?
Not according to the October 20 report. MiniMax said the weights would arrive on October 23, but the supplied information does not confirm that they were released or state what license will apply.
Source: ThorstenMeyerAI.com
