🔍 Read the full analysis: MiniMax M3.1 Flash: Ahead Of Open Models In Our Test, Behind On The Leaderboard on ThorstenMeyerAI.com
Get the latest gadgets 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, at $0.15 per million input tokens and $0.60 per million output tokens. It performed well in ThorstenMeyerAI.com’s 24-article newsroom test, but scored 43 on Artificial Analysis’s Intelligence Index, behind several open-weight models; the article does not provide the newsroom scores.
MiniMax released M3.1 Flash on October 19, pricing it at $0.15 per million input tokens and $0.60 per million output tokens. A review by ThorstenMeyerAI.com found that the model did well in its own 24-article newsroom test, while Artificial Analysis scored it 43 on its Intelligence Index, below several open-weight competitors. The difference matters for buyers weighing writing performance against broader benchmark standing, speed and cost.
MiniMax describes M3.1 Flash as a low-cost, fast model for agent work. The company says it is nearly three times faster than its M3 predecessor, with everyday speeds above 200 tokens per second and a median above 170. Artificial Analysis measured 120 tokens per second and a 1.08-second time to first token in its evaluation, below MiniMax’s stated speed figures. The company also says the new model costs about 70% less per agent task than M3; that figure is a MiniMax claim, not an independently established result in the review.
On Artificial Analysis’s Intelligence Index, M3.1 Flash scored 43. The index listed DeepSeek V4.1 Flash 1017 and GLM-5.3 at 45, Step 5 Preview at 44, and Kimi K3 at 44. The review also reports that M3.1 Flash used 180 million tokens to complete the index, compared with a median of 99 million. That is a raw token-use comparison, not a direct measure of total cost across every use case.
For the newsroom evaluation, ThorstenMeyerAI.com tested the model on 24 articles: 12 source items, each turned into an English and German article. Two judges, GPT-6.1 Sol and Claude Opus 5.5, scored the work against the source on fidelity, invented facts, clarity, usefulness, structure and polish. The supplied review material describes the setup but does not include the resulting model scores, so the size of its advantage over open-weight competitors cannot be verified from the published details provided 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
Different Tests Favor Different Strengths
The comparison shows why a single benchmark rank is an incomplete guide to a model’s usefulness. A newsroom task tests whether a model can turn specific source material into readable, accurate copy; an index combines a broader collection of evaluations. M3.1 Flash’s reported showing in the newsroom test and its 43-point index score describe different capabilities, not necessarily conflicting findings.
Price is another part of the decision. At its listed rates, M3.1 Flash is positioned as an inexpensive option, but the review found it consumed more tokens than the index median. Cost per token alone does not settle cost per completed task: output length, speed, quality and the amount of correction required also affect what a deployment costs. The source does not provide enough newsroom score data to calculate that trade-off.
The release also has practical relevance for developers considering a model for agent workflows. MiniMax is marketing M3.1 Flash for that use, but its claimed speed and task savings differ from what the independent index measured on speed. Teams will need to test their own workloads before treating either set of figures as a reliable forecast of performance in production.
As an affiliate, we earn on qualifying purchases.
From Space Bunny to M3.1 Flash
Before the formal announcement, the model ran anonymously on OpenRouter as Space Bunny Alpha. ThorstenMeyerAI.com reports that it appeared on September 23, led OpenRouter’s usage ranking for 10 days and processed 56.7 trillion tokens there before being removed on October 5. MiniMax says the model processed more than 200 trillion tokens overall, a company-reported total.
MiniMax said it revealed the model’s identity on October 19 and made M3.1 Flash available through its API, M Plan subscription, MiniMax Code and OpenRouter. OpenRouter offered free access for the first 72 hours, according to the review. The model supports a one-million-token context window; OpenRouter lists a maximum output of 524,288 tokens. The review says reasoning effort can be adjusted.
The model’s weights were not available when the review was published. MiniMax promised to release them on October 23. Until that happens, Artificial Analysis classifies the model as proprietary. MiniMax has not disclosed an official parameter count or a license. The review declines to use a parameter estimate reported by another outlet because MiniMax has not confirmed it.
““nearly three times faster” than M3”
— MiniMax
cost-effective AI text generation tools
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Scores, Weights and Data Handling
The review’s supplied material does not show the newsroom benchmark’s scores or the competing models’ results. It therefore establishes the test design, but not how far M3.1 Flash led, or whether the difference was large enough to matter in practice. The scores should not be inferred from the review’s headline.
MiniMax’s published benchmark charts compare the model with selected rivals, and the review says most are older versions. The results are company-reported, and MiniMax’s charts do not include DeepSeek in the Agents’ Last Exam comparison. The review also says DeepSeek’s model card reports a different score on that test, potentially using another setup. These figures are not directly comparable without matching evaluation conditions. The review says MiniMax’s security results, including its 82% CyberGym result, come from internal evaluation and have not been independently reproduced.
There is also an unresolved data question for people who used Space Bunny Alpha. According to the review, the provider could retain prompts during the anonymous period but could not use them for training. MiniMax has not said what happened to that retained data. The promised weight release, license terms and official parameter count were also outstanding in the source material.
As an affiliate, we earn on qualifying purchases.
Watch for the October 23 Release
The next stated milestone is MiniMax’s promised release of the M3.1 Flash weights on October 23, 2026. The release will clarify whether developers can inspect or run the model outside MiniMax’s hosted services, but the source material does not specify what license will apply. MiniMax has not announced an official parameter count.
Independent testing after the weights become available could help establish whether results from the hosted version transfer to other deployments. Users may also look for fuller reporting of the newsroom benchmark, including model-by-model scores, and for an explanation from MiniMax about data retained during the Space Bunny period. Until then, the strongest confirmed picture is mixed: a low listed price and a promising reported newsroom result, alongside a lower index score than several open-weight peers and speed measurements below the company’s claim.
open source AI models for news writing
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
How much does MiniMax M3.1 Flash cost?
MiniMax lists $0.15 per million input tokens and $0.60 per million output tokens. Cached input is listed at $0.015 per million tokens in the review.
Did M3.1 Flash beat open-weight models?
The review says it performed well in its own newsroom test, but the supplied material does not include scores or comparisons for that test. On Artificial Analysis’s Intelligence Index, it scored 43, behind several listed open-weight models.
How fast is the model?
MiniMax claims everyday speeds above 200 tokens per second and a median above 170. Artificial Analysis measured 120 tokens per second and 1.08 seconds to the first token in its evaluation.
Are the model weights available?
Not in the review’s account of the October 19 release. MiniMax promised the weights for October 23, 2026, but had not disclosed a license in the source material.
What happened to prompts sent to Space Bunny Alpha?
The review says the provider could retain prompts during the anonymous OpenRouter period but could not train on them. MiniMax has not explained what happened to retained data.
Source: ThorstenMeyerAI.com
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
