🔍 Read the full analysis: MiniMax M3.1 Flash Trails The Open-Weight Leaderboard, Then Surprises In Our Test 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 at $0.15 per million input tokens and $0.60 per million output tokens, with model weights promised for October 23. An independent index placed it behind several open-weight competitors, while ThorstenMeyerAI.com’s 24-article newsroom test found results that the source says complicate that picture; the supplied material does not include the test’s 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. Early evaluations present a mixed picture: the independent Artificial Analysis index scores it below several open-weight leaders, while ThorstenMeyerAI.com’s newsroom test says its results differed from those broader benchmarks; the supplied report excerpt does not include the test scores.
MiniMax offers the model through its API, M Plan subscription, MiniMax Code and OpenRouter. OpenRouter listed it free for the first 72 hours. The release includes a one-million-token context window, adjustable reasoning effort and a cached-input price of $0.015 per million tokens. MiniMax says it costs half as much per token as its M3 predecessor and about 70% less per agent task; those savings are company claims, not measurements established by the independent tests described in the report.
MiniMax claims everyday speeds above 200 tokens per second, with a median above 170. Artificial Analysis measured 120 tokens per second and 1.08 seconds to first token. Its index score for M3.1 Flash was 43, compared with 45 for DeepSeek V4.1 Flash 1017 and GLM-5.3, and 46 for Xiaomi’s MiMo-V2.6-Pro. Artificial Analysis also found the model used 180 million tokens in its index evaluation, against a 99-million-token median. That measurement suggests low per-token pricing does not by itself establish lower overall task cost.
MiniMax’s published benchmark charts show M3.1 Flash leading the listed models on Automation-Bench at 62.2, 1.7 points ahead of DeepSeek V4.1 Flash 1017. DeepSeek led on DeepSWE and Terminal-Bench. The company charts omit some rival results, and the report says MiniMax selected the competitors. A separate newsroom evaluation used 24 articles in English and German, based on 12 source materials and judged by GPT-6.1 Sol and Claude Opus 5.5. The available source text gives the test method but cuts off before providing scores or comparisons, so its outcome cannot be reported precisely.
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 Meets Real-World Performance
M3.1 Flash’s appeal rests on a combination of low listed prices and claims of high speed for agent tasks. But the assessments supplied do not establish that it is a leading model overall: Artificial Analysis placed it behind several competitors on its Intelligence Index, and measured lower throughput than MiniMax’s stated everyday figure. For users, the distinction matters because a cheap token can still add up if a model uses more tokens or requires extra work to produce a usable result.
The newsroom test targets a different question from broad benchmark suites: whether a model can turn source material into a factual article. That could be relevant to publishers considering models for production, but without the scores, the source excerpt does not support a conclusion about how M3.1 Flash performed against alternatives. The release is best read as a promising price-and-capability proposition with performance still needing independent, task-specific scrutiny.
As an affiliate, we earn on qualifying purchases.
From Space Bunny to M3.1 Flash
Before its public release, the model appeared anonymously on OpenRouter as Space Bunny Alpha. According to the report, it went live on September 23, led the platform’s usage ranking for 10 days and processed 56.7 trillion tokens there before being removed on October 5. MiniMax says it processed more than 200 trillion tokens overall. Those usage figures describe activity, not a public measure of model quality.
MiniMax identified Space Bunny Alpha as an early version of M3.1 Flash on October 19. The report says users’ prompts could be retained during the anonymous period, although the provider was not allowed to train on them. MiniMax has not clarified what happened to that data. The model’s weights are not yet available, so Artificial Analysis lists it as proprietary. MiniMax says the weights will arrive October 23; it has not published a license or confirmed an official parameter count.
“The weights “will arrive on Oct 23.””
— MiniMax
As an affiliate, we earn on qualifying purchases.
Weights, Data and Test Scores
Several points remain unresolved. The promised October 23 weight release had not happened at the time of the report, and MiniMax had not stated a license or official parameter count. A parameter estimate cited by another outlet was unconfirmed by the company and is not treated as established here.
The supplied source excerpt does not include the newsroom test’s results table, despite describing its 24-article method. It therefore does not show the model’s English or German scores, how it ranked against other models, or whether judges found differences in factual errors. MiniMax’s benchmark figures are self-reported, and the report says its security evaluations, including an 82% CyberGym result, came from an internal assessment that had not been independently reproduced. MiniMax also has not explained what happened to prompts retained during the Space Bunny period.
As an affiliate, we earn on qualifying purchases.
Watch for Weights and Scores
The next stated milestone is October 23, when MiniMax says it plans to publish M3.1 Flash’s weights. Their release, license terms and any confirmed model specifications will determine how developers can inspect or run the model outside MiniMax’s hosted services. The company has not supplied a later confirmation in the material available here.
Further evaluation should clarify how performance holds up across independent tests, including the newsroom benchmark once its missing scores are available. Buyers weighing the model against alternatives should compare task-level token use, speed and output quality, rather than relying on per-token price or company-selected benchmark results alone.
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, through its API, M Plan, MiniMax Code and OpenRouter. OpenRouter offered access free for the first 72 hours.
How much does M3.1 Flash cost?
The listed price is $0.15 per million input tokens, $0.60 per million output tokens and $0.015 per million cached input tokens. MiniMax says it costs half as much per token as M3, but actual task costs also depend on how many tokens a model uses.
How did M3.1 Flash compare with other models?
Artificial Analysis scored it 43 on its Intelligence Index, below MiMo-V2.6-Pro at 46 and DeepSeek V4.1 Flash 1017 and GLM-5.3 at 45. MiniMax’s own charts show M3.1 Flash ahead on Automation-Bench, while DeepSeek led on DeepSWE and Terminal-Bench.
Are the model weights available?
Not in the October 20 report. MiniMax said the weights would arrive on October 23, but the available material does not confirm their release or specify a license.
What did the newsroom test find?
The test evaluated 24 articles in English and German from 12 real news sources, with two AI judges scoring source fidelity, invented facts, clarity and other qualities. The supplied report excerpt omits the results table, so it does not establish the model’s scores or ranking.
Source: ThorstenMeyerAI.com
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
