🔍 Read the full analysis: MiniMax M3.1 Flash: Our Newsroom Test Compared It With Every Open Model on ThorstenMeyerAI.com
Get the little things that make your day 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 at $0.15 per million input tokens and $0.60 per million output tokens. An independent index measured lower speed and intelligence scores than some open-weight rivals, while a newsroom test covered 24 articles; the supplied source does not include that test’s results. The model’s weights are scheduled for October 23, but had not been released when the source was published.
MiniMax released M3.1 Flash on October 19, pricing the model at $0.15 per million input tokens and $0.60 per million output tokens as it targets agent tasks. A review by ThorstenMeyerAI.com compared the company’s claims with Artificial Analysis measurements and a 24-article newsroom test; the supplied review excerpt gives the test setup but not its scores, leaving its editorial performance results unavailable.
The model is available through the MiniMax API, M Plan subscription, MiniMax Code and OpenRouter. OpenRouter offered access free for the first 72 hours after launch. Cached input is priced at $0.015 per million tokens. The listed input and output rates are half the corresponding rates for MiniMax M3, according to the source. MiniMax also says M3.1 Flash costs about 70% less per agent task, a claim that depends on task-level usage rather than token prices alone.
MiniMax advertises a one-million-token context window and says the model can exceed 200 tokens per second in everyday use, with a median above 170. Artificial Analysis measured 120 output tokens per second and 1.08 seconds to the first token in its evaluation. That is below the company’s stated speed figures and below the 211 tokens per second it measured for DeepSeek V4.1 Flash 1017. The review says MiniMax described the model’s design only as using “more aggressive sparsity.”
The review also reports that M3.1 Flash was previously available anonymously on OpenRouter as Space Bunny Alpha. It went live on September 23, led OpenRouter’s usage ranking for 10 days and processed 56.7 trillion tokens there before being withdrawn on October 5, according to the source. MiniMax says the model processed more than 200 trillion tokens overall. The model’s weights had not been released as of the review; MiniMax said they would arrive October 23.
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 Versus Measured Performance
M3.1 Flash’s low listed price may make it attractive for developers running high volumes of agent requests, but price per token does not establish cost per completed task. Artificial Analysis found that the model used 180 million tokens to complete its Intelligence Index evaluation, compared with a median of 99 million. That is a separate measure from the cost of a specific agent workflow, but it shows why token efficiency can change the practical value of a low rate.
The independent index gave M3.1 Flash an Intelligence Index score of 43. In the comparison cited by the review, MiMo-V2.6-Pro scored 46, while DeepSeek V4.1 Flash 1017 and GLM-5.3 scored 45 each. The results place MiniMax’s model among strong open-weight competitors, but not at the top of that group on this index. Artificial Analysis listed it as proprietary at the time because its weights were not yet available.
For newsrooms and other users who need source-grounded writing, broad benchmark scores are not a substitute for testing factual accuracy and instruction-following on real assignments. The review’s 24-article setup is relevant to that use case, but without the reported scores readers cannot tell how M3.1 Flash performed against the other models tested.
As an affiliate, we earn on qualifying purchases.
From Anonymous Traffic to Release
Before MiniMax identified the model, Space Bunny Alpha appeared on OpenRouter on September 23, 2026, and remained there until October 5. The source says users could access it anonymously during that period and that MiniMax was permitted to retain prompts, though not to use them for training. MiniMax had not clarified what happened to data retained during the anonymous period, according to the review.
MiniMax’s published benchmark charts show M3.1 Flash scoring 62.2 on Automation-Bench, compared with 60.5 for DeepSeek V4.1 Flash 1017. On DeepSWE, DeepSeek scored 77.0 to MiniMax’s 73.7, and on Terminal-Bench DeepSeek scored 42.3 to 38.1. The review notes that MiniMax selected the models in its comparison and that its figures are company-reported; it says the security results, including an 82% CyberGym score, came from an internal evaluation and had not been independently reproduced.
MiniMax described M3.1 Flash as “first among Chinese models” on Agents’ Last Exam. The review cautions that its chart did not include DeepSeek in that row, while DeepSeek’s model card reported a score of 35.4, potentially under a different setup. The available material does not establish a like-for-like comparison.
““More than 200 tokens per second” in everyday use, with a median above 170.”
— MiniMax
As an affiliate, we earn on qualifying purchases.
Newsroom Scores Still Missing
The supplied source excerpt describes a test of 12 real news sources in English and German, producing 24 articles, but ends before presenting the results table. It says each article was scored by GPT-6.1 Sol and Claude Opus 5.5 for source fidelity, invented facts, clarity, usefulness, structure and polish. Without the scores or comparisons, no conclusion about M3.1 Flash’s newsroom quality can be drawn from the material provided.
Other open questions include whether the promised weights will be released on October 23, what license will govern them, and how many parameters the model has. MiniMax has not confirmed a parameter count; the review declines to rely on an unconfirmed figure reported elsewhere. The company also has not clarified what happened to prompts retained during the Space Bunny Alpha period.
Finally, MiniMax’s benchmark figures and task-cost estimate remain company claims, while the review says its security results have not been independently reproduced. Differences in benchmark setup may also affect comparisons, including the Agents’ Last Exam result.
As an affiliate, we earn on qualifying purchases.
Weights and Results Due Next
The next stated milestone is October 23, 2026, when MiniMax says it plans to release M3.1 Flash’s weights. If released, the weights and license would allow developers to assess how the model can be deployed and independently examined. The company had not published an official parameter count in the material available to the review.
Readers seeking a verdict on newsroom use will need the missing scores from the 24-article test, including comparisons with the other models tested. Further independent evaluations could also clarify whether MiniMax’s speed and cost claims hold across tasks, and whether the performance measured by Artificial Analysis changes after the weights become available.
AI token usage monitoring software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
When did MiniMax release M3.1 Flash?
MiniMax announced and released M3.1 Flash on October 19, 2026, through its API, M Plan, MiniMax Code and OpenRouter, according to the source.
How much does M3.1 Flash cost?
The listed price is $0.15 per million input tokens and $0.60 per million output tokens. Cached input is listed at $0.015 per million tokens.
Did M3.1 Flash outperform other open-weight models?
Not across the comparisons in the supplied material. Artificial Analysis scored it at 43, below several listed rivals, and measured it at 120 output tokens per second. MiniMax’s own charts show it ahead of DeepSeek V4.1 Flash 1017 on Automation-Bench, while DeepSeek scored higher on DeepSWE and Terminal-Bench.
What did the newsroom test find?
The source describes a 24-article test using 12 news sources in English and German, with two judges rating six writing dimensions. The excerpt does not include the results, so its findings cannot be reported from the supplied material.
Are the model weights available?
They had not been released when the source review was published. MiniMax said the weights would arrive October 23, 2026, but the source does not confirm that release or specify a license.
Source: ThorstenMeyerAI.com
Fall Picks
fall essentials
As an affiliate, we earn on qualifying purchases.
