🔍 Read the full analysis: MiniMax M3.1 Flash: A Closer Look At Its Open-Model Test Results on ThorstenMeyerAI.com
Get audio and creator gear delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
TL;DR
MiniMax released M3.1 Flash on October 19, 2026, pricing it below its predecessor and promoting it for agent tasks. Its own benchmarks show strengths in automation, while Artificial Analysis rates it behind several open-weight rivals on intelligence and speed. The supplied source does not include the newsroom test scores, so those results cannot be assessed here; the model weights and license are also still pending.
MiniMax released M3.1 Flash on October 19, marketing it as a low-cost, fast model for agent tasks, but the available test results present a mixed picture: the company reports a strong automation score, while an independent index places it behind several open-weight competitors. The model is available through MiniMax’s API and other services, but its weights have not yet been released, limiting independent inspection and meaning it is currently listed as proprietary by Artificial Analysis.
MiniMax lists the price at $0.15 per million input tokens and $0.60 per million output tokens, with cached input at $0.015 per million. OpenRouter lists the preceding M3 model at twice the input and output rates. MiniMax says M3.1 Flash is nearly three times faster than M3 in everyday use, with more than 200 tokens per second and a median above 170. Those are company claims, not results from the independent index.
MiniMax’s published benchmark chart gives M3.1 Flash a 62.2 on Automation-Bench, ahead of the listed rivals, including DeepSeek V4.1 Flash 1017 at 60.5. But DeepSeek leads the chart’s DeepSWE coding result, 77.0 to 73.7, and Terminal-Bench, 42.3 to 38.1. The chart does not report every competitor on every test. MiniMax selected the comparison models, and its figures have not been independently reproduced, according to the source review.
Artificial Analysis assigns M3.1 Flash an Intelligence Index score of 43. The review lists several open-weight competitors at 44 or higher, including DeepSeek V4.1 Flash 1017 and GLM-5.3 at 45. It measured M3.1 Flash at 120 tokens per second and 1.08 seconds to first token—below MiniMax’s speed claim and below DeepSeek’s measured 211 tokens per second. The source also reports that M3.1 Flash used 180 million tokens in the index evaluation, compared with a 99 million-token median.
The review says its newsroom assessment covered 24 articles: 12 sources rendered in both English and German, with two independent judges scoring the results. However, the supplied source material ends before giving the scores or findings. Those results cannot be reported or compared from the material available here.
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, Speed and Benchmark Trade-offs
M3.1 Flash’s low listed price could matter to developers running agents at scale, where repeated model calls can make token costs add up. But price per token is only part of the cost: the source review says the model used 180 million tokens in Artificial Analysis’s evaluation, against a 99 million median. If that observation holds across practical workloads, greater token use could offset some of the lower unit price. The figures do not establish how costs compare for every task.
The results also caution against treating one benchmark as a complete measure of capability. M3.1 Flash leads the company’s listed Automation-Bench comparison, but trails DeepSeek on two coding tests and several models on the independent index. For buyers, the relevant choice depends on their task, output quality, latency and total token use—not only a headline score or vendor pricing.
Open weights would allow researchers and organizations to inspect and run the model outside MiniMax’s hosted services, subject to its eventual license. Until they are released, the model’s parameters and licensing terms remain unavailable in the source material. That leaves an important part of the “open-model” comparison unresolved.
As an affiliate, we earn on qualifying purchases.
From Space Bunny to M3.1 Flash
Before MiniMax identified it, the model appeared anonymously on OpenRouter under the name Space Bunny Alpha. The source review says it went live on September 23, led the platform’s usage ranking for 10 days, and was removed on October 5, after processing 56.7 trillion tokens there. MiniMax says total usage across platforms exceeded 200 trillion tokens. These figures are attributed to the review and company, respectively.
MiniMax revealed the model’s identity on October 19 and made it available through its API, M Plan subscription, MiniMax Code and OpenRouter. OpenRouter offered it free for its first 72 hours. MiniMax says the model uses “more aggressive sparsity,” but the source provides no further design details. It supports a one-million-token context window; OpenRouter lists a maximum output of 524,288 tokens, and reasoning effort is adjustable.
MiniMax says the weights will arrive on October 23. The source review notes that the company has not disclosed an official parameter count or license. A reported parameter estimate of 428 billion total and 23 billion active is explicitly unconfirmed by MiniMax and is not treated here as established fact.
““nearly three times faster” than M3”
— MiniMax
Weights, Data and Test Results Pending
The source does not establish whether the weights will appear on October 23, what license will govern them, or how many parameters the model has. MiniMax’s benchmark figures are self-reported, and the source says they have not yet been reproduced independently. Its comparison also omits some rivals from individual rows; for example, MiniMax’s chart does not list DeepSeek in the Agents’ Last Exam row, while the source review cites a separate DeepSeek model-card score of 35.4 and warns that the setups may differ.
There is also a data-retention question related to Space Bunny Alpha. According to the review, during the anonymous period the provider could retain prompts but could not use them for training. MiniMax has not said what happens to that retained data. The supplied material also omits the newsroom benchmark’s scores and conclusions, so no claim about the model’s performance on those 24 articles can be made from this source.
The October 23 Weights Release
The next stated milestone is October 23, 2026, when MiniMax says it plans to release the weights. Their arrival, license and technical documentation will determine whether researchers and developers can run and inspect the model independently. A release would also allow more direct comparisons than are possible while the model is available only through hosted services.
Readers should also look for independently repeated benchmark results and clarification from MiniMax about prompts retained during the anonymous OpenRouter period. The newsroom test scores are not present in the supplied source text; without them, that portion of the review remains unresolved. Until these details are available, the evidence supports a limited conclusion: M3.1 Flash is competitively priced, but its measured speed and benchmark standing vary by evaluator and task.
Key Questions
When was MiniMax M3.1 Flash released?
MiniMax revealed and released M3.1 Flash on October 19, 2026, through its API and several services, including OpenRouter.
How much does M3.1 Flash cost?
MiniMax lists prices of $0.15 per million input tokens and $0.60 per million output tokens. Cached input is listed at $0.015 per million tokens.
Are the model weights available?
Not according to the supplied source, which says MiniMax promised the weights for October 23, 2026. The license and official parameter count were not disclosed in that material.
How did it perform against competing models?
The results are mixed. MiniMax’s chart shows M3.1 Flash leading its listed rivals on Automation-Bench, while DeepSeek V4.1 Flash 1017 scores higher on DeepSWE and Terminal-Bench. Artificial Analysis gives M3.1 Flash an Intelligence Index score of 43, below several listed competitors.
What did the 24-article newsroom test find?
The supplied source describes a test of 12 news sources in English and German, scored by two judges, but it does not include the results. No conclusion about those scores can be drawn from the material provided.
Source: ThorstenMeyerAI.com
Fall Picks
fall essentials
As an affiliate, we earn on qualifying purchases.
