🔍 Read the full analysis: Could MiniMax M3.1 Flash Be Better Than Its Leaderboard Rank Suggests? on ThorstenMeyerAI.com
Get everyday essentials 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 at $0.15 per million input tokens and $0.60 per million output tokens, pitching it as a fast model for agent tasks. An independent index scores it below several open-weight competitors, while the source review’s 24-article newsroom test setup is described but its results are missing. The weights are expected October 23, but their license and other technical details have not been disclosed.
MiniMax released M3.1 Flash on October 19, pricing it at $0.15 per million input tokens and $0.60 per million output tokens and promoting it for agent tasks. But the available evidence does not establish that it outperforms its leaderboard rank: Artificial Analysis gives it an Intelligence Index score of 43, behind several open-weight models, and the supplied review material does not include the results of its separate newsroom test.
The model is available through the MiniMax API, M Plan subscription, MiniMax Code and OpenRouter. OpenRouter offered access free for its first 72 hours. The listed price is half the per-token price of MiniMax M3, according to the review; MiniMax also says M3.1 Flash costs about 70% less per agent task. That task-cost comparison is a company claim, not an independently verified result in the material provided.
MiniMax claims everyday output speeds above 200 tokens per second, with a median above 170, and says the model is “nearly three times faster” than M3. Artificial Analysis measured 120 tokens per second and 1.08 seconds to first token. The independent result is lower than the company’s stated everyday speed. MiniMax has described the architecture only as using “more aggressive sparsity.” It lists a one-million-token context window; OpenRouter allows up to 524,288 output tokens, and reasoning effort can be adjusted.
The model had previously appeared anonymously on OpenRouter as Space Bunny Alpha. The review says it went live September 23, led OpenRouter’s usage ranking for 10 days, processed 56.7 trillion tokens on the platform, and was removed October 5. MiniMax says total use across services exceeded 200 trillion tokens. Those figures describe platform usage, not a measure of model quality.
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 Real-World Performance
M3.1 Flash’s low listed price may make it attractive for applications that send many requests, including agent workflows. But per-token price is only part of operating cost. Artificial Analysis found that the model used 180 million tokens to complete its index evaluation, compared with a median of 99 million. A model that consumes more tokens can narrow the savings from a low rate, depending on the task and how it is deployed.
The performance picture is mixed rather than settled. M3.1 Flash’s Intelligence Index score of 43 is below scores of 45 and 46 for several models in the comparison, while its measured speed is below DeepSeek V4.1 Flash 1017’s 211 tokens per second. MiniMax’s own benchmark chart reports a lead on Automation-Bench, but DeepSeek leads on DeepSWE and Terminal-Bench. Different evaluations measure different capabilities, so a single rank cannot answer whether the model is better for a particular use.
The review also says it tested the model on 24 articles based on 12 real news sources, producing English and German versions and using two independent judges. However, the source material ends before giving any scores or comparative results. That means readers cannot use this account to assess whether M3.1 Flash was accurate, clear or prone to invented details in that test.
As an affiliate, we earn on qualifying purchases.
From Space Bunny to M3.1 Flash
MiniMax identified Space Bunny Alpha as an early version of M3.1 Flash after its anonymous run on OpenRouter. The review reports that the model ranked as the platform’s most-used model for 10 days in late September. That usage history drew attention to the model before MiniMax’s October 19 release, but popularity on a platform is not the same as independent evidence of accuracy or capability.
MiniMax’s published comparison includes Automation-Bench, DeepSWE, Terminal-Bench, OSWorld V2, Agents’ Last Exam and HLE. The reviewer says the company selected the competitors and that many listed rivals were older versions, including GPT-5.6 Sol and Claude Opus 5 rather than newer versions named in the source. The scores are self-reported by MiniMax, and the review says they have not yet been independently reproduced. Some cells are marked “not reported,” limiting direct comparisons.
Artificial Analysis’s comparison places M3.1 Flash among leading open-weight models but below several listed competitors. It currently categorizes M3.1 Flash as proprietary because its weights have not been released. MiniMax says the weights will arrive October 23; it has not supplied a license or confirmed a parameter count. A separate outlet reportedly gave a parameter estimate, but MiniMax has not confirmed it, so that figure is not treated here as established.
““nearly three times faster” than M3”
— MiniMax
As an affiliate, we earn on qualifying purchases.
Benchmark Results Still Have Gaps
Several questions remain open. The supplied review describes its newsroom test but provides no scores, model comparisons or findings from the judges. It is therefore not possible to say from this material whether M3.1 Flash performed well on source fidelity, invented facts, clarity or structure. The test used a free OpenRouter endpoint and a 32,000-token output limit, but the missing results prevent readers from evaluating what happened.
MiniMax’s benchmark results have not been independently reproduced according to the reviewer. The company’s chart omits DeepSeek from its Agents’ Last Exam row, while DeepSeek’s model card reports a different score; the review notes that the setups may differ. MiniMax’s security results, including an 82% score on CyberGym, are described as coming from an internal evaluation. The source does not provide an independent confirmation.
MiniMax has also not disclosed the weights’ license or confirmed the model’s parameter count. The review says providers could retain prompts during Space Bunny Alpha’s anonymous period, though they were not allowed to train on them. MiniMax has not said what happened to that stored data, leaving the handling of prompts from that period unresolved.
As an affiliate, we earn on qualifying purchases.
Weights and Test Data to Watch
The next stated milestone is October 23, when MiniMax says it plans to release M3.1 Flash’s weights. Their publication would allow observers to check the model’s technical details and determine what license governs use. The source material does not confirm whether the release has occurred or whether the date has changed.
For a clearer performance comparison, the missing newsroom scores and the exact benchmark setups matter. Independent replications of MiniMax’s chart, including tests against current competitors under matched conditions, could clarify where M3.1 Flash excels and where it falls short. Until then, the available evidence supports a measured conclusion: the model is inexpensive by its listed token rates, but the case that it is better than its independent rank suggests remains unproven in the supplied reporting.
As an affiliate, we earn on qualifying purchases.
Key Questions
What is MiniMax M3.1 Flash?
It is a MiniMax model released October 19, 2026, and promoted for agent tasks. It is available through the MiniMax API, M Plan, MiniMax Code and OpenRouter.
How much does M3.1 Flash cost?
The listed rate is $0.15 per million input tokens, $0.60 per million output tokens and $0.015 per million cached input tokens. OpenRouter offered the model free for its first 72 hours.
Does independent testing show it is faster than rivals?
Not across the comparisons provided. Artificial Analysis measured 120 tokens per second, below MiniMax’s claim of more than 200 in everyday use and below DeepSeek V4.1 Flash 1017’s listed 211 tokens per second.
What did the 24-article newsroom test find?
The supplied material describes a test using 12 sources in English and German, scored by two judges, but does not include the scores or conclusions. No result from that test can be established from the text provided.
When will the model weights be available?
MiniMax said the weights would arrive October 23, 2026. The supplied material does not confirm publication, and it says the license and parameter count have not been disclosed by MiniMax.
Source: ThorstenMeyerAI.com
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
