🔍 Read the full analysis: MiniMax M3.1 Flash: 3 Things Our Newsroom Test Showed About Its Performance on ThorstenMeyerAI.com
Get hardware and tech 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, 2026, pricing it at $0.15 per million input tokens and $0.60 per million output tokens. The supplied source reports mixed results across MiniMax’s own benchmarks and an independent index, but cuts off before giving the newsroom test’s scores, so that test’s findings cannot be reported here.
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. A report dated October 20 says the model’s performance looks mixed across company-reported benchmarks and an independent index. The report also describes a 24-article newsroom test, but the supplied material ends before its scores or conclusions, leaving that part of the performance picture unconfirmed.
MiniMax presents M3.1 Flash as a fast, lower-cost model for agent work. The listed price is half the per-token rate of its predecessor, M3, according to the report. MiniMax also claims the new model costs about 70% less per agent task. That comparison is the company’s claim; the source does not provide the task conditions or an independent cost test to verify it.
The report says MiniMax claims everyday generation speeds above 200 tokens per second, with a median above 170, and describes the model as nearly three times faster than M3. Artificial Analysis, an independent evaluator cited in the report, measured 120 tokens per second and 1.08 seconds to first token. That result is below MiniMax’s stated speed claim and below the reported 211 tokens per second measured for DeepSeek V4.1 Flash 1017.
The independent index gave M3.1 Flash an Intelligence Index score of 43. The report lists several open-weight competitors above it, including MiMo-V2.6-Pro at 46 and DeepSeek V4.1 Flash 1017 and GLM-5.3 at 45. Artificial Analysis also found the model used 180 million tokens during its index evaluation, compared with a 99 million median. Those measurements suggest that its low price per token may not tell the whole cost story if a task uses more tokens.
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 appeal depends on more than its headline price. For developers running agent workflows, the relevant costs include tokens consumed, response speed and task quality, as well as the listed rate per million tokens. The independent evaluator’s reported token use and speed provide a caution against treating a lower per-token price as proof of a lower total cost or faster output.
The available benchmark results also show why comparisons need care. In MiniMax’s published charts, M3.1 Flash scored 62.2 on Automation-Bench, ahead of DeepSeek V4.1 Flash 1017’s 60.5. DeepSeek led on DeepSWE and Terminal-Bench, however, and MiniMax’s charts did not report every rival’s score on every test. These are company-selected comparisons, not a complete independent ranking.
The newsroom test could have helped readers judge a practical use case: whether the model can turn source material into accurate articles without adding unsupported details. The source describes the test design but provides no scores, model comparisons or newsroom findings. It cannot support a conclusion about whether M3.1 Flash performed well on that task.
AI language model performance testing tools
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
From Space Bunny to M3.1
Before its public identification, the model appeared on OpenRouter under the name Space Bunny Alpha. The report says it went live anonymously 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. The source does not independently verify those figures.
MiniMax identified Space Bunny Alpha as an early version of M3.1 Flash on October 19. The model is available through the MiniMax API, M Plan subscription, MiniMax Code and OpenRouter, according to the report; OpenRouter offered it free for its first 72 hours. It has a 1-million-token context window. The report says weights were promised for October 23, but they had not been released at the time covered by the source.
MiniMax has not published an official parameter count or license in the supplied material. A parameter estimate attributed to another outlet is explicitly unconfirmed and is not used here. Artificial Analysis consequently lists M3.1 Flash as a proprietary model while its weights remain unavailable.
As an affiliate, we earn on qualifying purchases.
Missing Newsroom Test Scores
The source material does not include the results of the newsroom benchmark. It describes 12 real news sources written in English and German, producing 24 articles, and says two judges scored fidelity, invented facts, clarity, usefulness, structure and polish. But the table begins and then cuts off before any scores appear. The test’s findings, rankings and any comparison with other models are therefore unknown from the material provided.
Other points also remain unsettled. MiniMax’s speed and agent-cost claims have not been independently confirmed in the source, and the company’s benchmark results have not been reproduced independently, according to the report. The source also says MiniMax has not explained what happened to prompts retained during the anonymous Space Bunny period. It reports that the provider could keep prompts but could not use them for training; MiniMax’s handling of that data remains unclear.
As an affiliate, we earn on qualifying purchases.
Weights and Results to Watch
The next stated milestone in the source is the planned October 23 release of M3.1 Flash’s weights. Their publication would allow outside researchers to inspect and test the model, though the source does not specify a license or confirm that the promised release occurred. Readers should also look for the complete newsroom benchmark table before drawing conclusions about performance on article-writing tasks.
Further independent testing could clarify how M3.1 Flash compares with competing models on speed, token use, accuracy and total task cost. Until those results are available, the reported index score and company-selected benchmarks provide only part of the picture. As with other AI services, performance can vary by prompt and task, and API costs may change with usage.
As an affiliate, we earn on qualifying purchases.
Key Questions
When did MiniMax release M3.1 Flash?
MiniMax announced the model on October 19, 2026, according to the source report dated October 20.
How much does M3.1 Flash cost?
The reported API rates are $0.15 per million input tokens and $0.60 per million output tokens. Cached input is listed at $0.015 per million tokens. Actual spending depends on how much a task uses.
Did the newsroom test show that M3.1 Flash writes accurate articles?
The supplied source describes a 24-article test but does not include its scores or conclusions. It is not possible to say from this material whether the model performed well on article accuracy or other newsroom criteria.
Is M3.1 Flash open-weight?
Not according to the status described in the report: its weights had not yet been released, and MiniMax had said they were expected on October 23. The source does not confirm whether that release later happened.
How does its measured speed compare with MiniMax’s claim?
MiniMax claimed everyday speed above 200 tokens per second. Artificial Analysis measured 120 tokens per second, according to the report. The two figures come from different measurement sources and should not be treated as directly equivalent without further test details.
Source: ThorstenMeyerAI.com
Evergreen bestsellers Picks
bestsellers
As an affiliate, we earn on qualifying purchases.
