🔍 Read the full analysis: MiniMax M3.1 Flash: The Leaderboard Says One Thing, Our Test Another on ThorstenMeyerAI.com
Get smart everyday buys 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. MiniMax’s own benchmark charts show strengths on some tasks, while Artificial Analysis measured a 43 Intelligence Index score and 120 tokens per second; a separate 24-article newsroom test is described, but its results are not included in the available source material.
MiniMax released M3.1 Flash on October 19, offering it through its API and several services at $0.15 per million input tokens and $0.60 per million output tokens. A review by ThorstenMeyerAI.com found that MiniMax’s claims, an independent model index and the outlet’s newsroom evaluation did not present a single consistent picture of the model’s performance; the available source describes the newsroom test but does not provide its scores.
MiniMax positions M3.1 Flash as a low-cost model for agent tasks. Its listed price is half the per-token price of the earlier M3, while MiniMax says the new model costs about 70% less per agent task. The company also claims everyday speed above 200 tokens per second and a median above 170, describing it as nearly three times faster than M3. Those speed and cost comparisons are MiniMax’s claims, not results established by the independent measurements cited in the review.
The model has a one-million-token context window and adjustable reasoning effort. Artificial Analysis measured its output speed at 120 tokens per second, with 1.08 seconds to the first token. That is below MiniMax’s stated everyday-speed figure and below the 211 tokens per second reported by Artificial Analysis for DeepSeek V4.1 Flash 1017. The independent index gave M3.1 Flash an Intelligence Index score of 43, compared with scores of 45 for DeepSeek V4.1 Flash 1017 and GLM-5.3, and 46 for Xiaomi’s MiMo-V2.6-Pro.
MiniMax’s published benchmark charts show M3.1 Flash leading on Automation-Bench, with 62.2 against DeepSeek’s 60.5, but DeepSeek scored higher on DeepSWE and Terminal-Bench. MiniMax’s chart gives M3.1 Flash 73.7 on DeepSWE and 38.1 on Terminal-Bench; DeepSeek scored 77.0 and 42.3 respectively. The figures are company-reported results, and the review says they have not been independently reproduced.
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 enters a market where buyers weigh per-token pricing against task performance. A lower price can make a model attractive for high-volume agent workloads, but speed and the number of tokens needed to complete a task also affect practical cost. Artificial Analysis found that M3.1 Flash used 180 million tokens during its index evaluation, compared with a 99 million-token median. That is an observation from that evaluation, not proof that every user or task will consume more tokens.
The results also caution against treating one benchmark or company ranking as a complete verdict. MiniMax’s charts show wins and losses across different tasks, while the independent index places the model behind several open-weight competitors. For teams choosing a model, the relevant question is whether it performs reliably on their own tasks at an acceptable total cost—not simply whether its advertised token price is low.
As an affiliate, we earn on qualifying purchases.
From Space Bunny to M3.1 Flash
Before its public launch, the model ran anonymously on OpenRouter as Space Bunny Alpha. The source says it appeared on September 23, led OpenRouter’s usage ranking for 10 days and processed 56.7 trillion tokens on the platform before being removed on October 5. MiniMax says the model processed more than 200 trillion tokens overall. These figures describe reported platform use; they do not establish how the model performs on a specific evaluation.
MiniMax revealed the identity behind Space Bunny on October 19 and released M3.1 Flash through the MiniMax API, M Plan subscription, MiniMax Code and OpenRouter. OpenRouter offered access free for the first 72 hours, according to the source. The model’s weights were not yet available when the review was published; MiniMax said they would arrive October 23. Until then, Artificial Analysis listed the model as proprietary.
MiniMax has not confirmed a parameter count or provided a license in the material cited. The review notes that one outlet reported 428 billion total parameters and 23 billion active parameters, but says it did not use those figures because MiniMax had not confirmed them.
“More than 200 tokens per second in everyday use and a median above 170.”
— MiniMax
low-cost AI chatbot development tools
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Newsroom Scores and Data Handling
The review says it tested M3.1 Flash on 24 news articles: 12 source texts each produced in English and German, scored by GPT-6.1 Sol and Claude Opus 5.5. However, the supplied source material ends before giving the results table. It is therefore not possible to report the model’s newsroom scores or say whether it outperformed other models on source fidelity, factual accuracy, clarity or structure.
Other points remain unsettled. MiniMax has not confirmed the model’s parameter count or announced a license in the cited material, and its benchmark figures have not been independently reproduced. The source also says OpenRouter permitted the provider to retain prompts during the Space Bunny period, while prohibiting their use for training. MiniMax has not said what happened to that data. The review provides no further detail on retention or deletion.
tokens per second AI processing hardware
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Weights and Independent Testing
The next stated milestone is the planned October 23 release of M3.1 Flash’s weights. If released as MiniMax says, the weights could allow researchers and developers to examine and test the model outside the company’s hosted services; the applicable license and exact release terms were not stated in the source.
Further independent evaluations could clarify how the model performs across coding, agent and newsroom tasks, and whether its speed and token use vary substantially by workload. Readers should also look for the missing newsroom-test results and any response from MiniMax about prompt data from the anonymous OpenRouter period. Until those details are available, the evidence supports a mixed assessment rather than a definitive ranking.
As an affiliate, we earn on qualifying purchases.
Key Questions
When did MiniMax release M3.1 Flash?
MiniMax released M3.1 Flash on October 19, 2026, according to the source. It was available through the MiniMax API, M Plan, MiniMax Code and OpenRouter.
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 lead the independent index?
No. Artificial Analysis gave it an Intelligence Index score of 43. The cited index listed MiMo-V2.6-Pro at 46 and DeepSeek V4.1 Flash 1017 and GLM-5.3 at 45.
What did the 24-article newsroom test find?
The source describes the test setup but does not include its results table. It says 12 real news sources were each turned into English and German articles and evaluated by two judges. No newsroom score can be confirmed from the supplied material.
Are M3.1 Flash’s weights available?
They had not been released when the review was published. MiniMax said the weights would arrive October 23; the source does not confirm whether that later release occurred.
Source: ThorstenMeyerAI.com
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
