🔍 Read the full analysis: MiniMax M3.1 Flash: An Open-Model Test Beyond The Leaderboard 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; its weights are promised for October 23. The source report finds strong value but says independent speed and intelligence measurements trail some open-weight rivals. It describes a 24-article newsroom test, but the supplied material cuts off before giving the results.
M3.1 Flash is available through the MiniMax API, M Plan subscription, MiniMax Code and OpenRouter. OpenRouter offered it free for its first 72 hours. The listed price for cached input is $0.015 per million tokens. OpenRouter lists the previous MiniMax M3 at $0.30 per million input tokens and $1.20 per million output tokens, making the new model’s per-token rates half those figures.
MiniMax describes the model as suited to fast, low-cost agent work. The company claims everyday speed above 200 tokens per second, with a median above 170, and says it is nearly three times faster than M3. Artificial Analysis measured 120 tokens per second and a 1.08-second time to first token. Those are measurements from a separate evaluation, not a confirmation of MiniMax’s claimed everyday speed.
The source report says the model supports a one-million-token context window and adjustable reasoning effort. It also reports that MiniMax has not published the model’s parameter count or a license. The weights had not been released as of the report’s October 20 publication date; MiniMax said they would arrive October 23.
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 Measured Performance
The release puts a lower-priced model into a fast-moving field where buyers weigh cost per token, speed and task quality. But a lower listed rate alone does not establish lower overall costs: Artificial Analysis found M3.1 Flash used 180 million tokens for its index evaluation, compared with a 99-million-token median. That difference matters because a model that consumes more tokens can narrow the savings from cheaper input and output.
The report’s benchmark comparisons also point to a trade-off rather than a clear lead. In MiniMax’s own chart, M3.1 Flash scored 62.2 on Automation-Bench, ahead of the listed rivals, but DeepSeek V4.1 Flash 1017 scored higher on DeepSWE and Terminal-Bench. Artificial Analysis gave MiniMax a 43 on its Intelligence Index, below several open-weight models in its comparison. These tests measure different tasks and should not be treated as a single ranking.
For teams considering a deployment, the unreleased weights also matter. Until they are available under a known license, users cannot assess the model’s actual openness or independently inspect and run the weights. The announced release date is a company commitment, not confirmation that publication has occurred.
As an affiliate, we earn on qualifying purchases.
From Space Bunny to M3.1
Before its public release, the model ran anonymously on OpenRouter as Space Bunny Alpha. According to Meyer’s report, it appeared on September 23, led OpenRouter’s usage ranking for 10 days and processed 56.7 trillion tokens there before being pulled on October 5. MiniMax says total usage across platforms exceeded 200 trillion tokens.
The report says the anonymous provider could retain prompts during that period, although it was not allowed to use them for training. MiniMax has not said what happened to that data. That unresolved question is separate from the model’s public release and its later performance testing.
MiniMax published benchmark comparisons as chart images, which the report says CellCog transcribed into a table. The company selected the comparison models, and the report notes that several were older versions than those available at the time. MiniMax’s cited security results, including an 82% CyberGym score, came from an internal evaluation, according to the source; independent replication was not yet available.
““more than 200 tokens per second” in everyday use, with a median above 170”
— MiniMax
As an affiliate, we earn on qualifying purchases.
Open Weights and Test Results
The supplied source material ends after introducing the newsroom test and a table heading, before giving any scores or findings. It therefore does not support a conclusion about how M3.1 Flash performed on the 24 English- and German-language articles, or how it compared with other models on that test.
Other open questions include whether the promised weights were released on October 23, what license they use, and whether the anonymous Space Bunny period’s retained prompts were deleted or otherwise handled. The source also says MiniMax had not confirmed a parameter count. Its company-selected benchmark results and internal security scores had not been independently reproduced in the report.
MiniMax’s claim that M3.1 Flash ranked first among Chinese models on Agents’ Last Exam also needs qualification: its chart did not include DeepSeek on that row, while DeepSeek’s model card reported a score of 35.4, potentially under a different setup. The available information does not establish a like-for-like comparison.
As an affiliate, we earn on qualifying purchases.
The October 23 Weights Release
The next stated milestone is MiniMax’s planned October 23 release of the weights. Their publication, license and availability will determine whether developers can run the model independently and verify its architecture and performance. The source material does not confirm that this release has happened.
A fuller assessment also depends on the missing newsroom benchmark results and independent testing of the company’s speed, agent-cost and security claims. Until then, readers can compare the reported price and available evaluation scores, but should not treat the incomplete evidence as proof of real-world performance or savings.
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 October 20 report. The source says it was offered through the MiniMax API, M Plan subscription, MiniMax Code and OpenRouter.
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. OpenRouter offered the model free for its first 72 hours.
Are M3.1 Flash’s weights available?
They had not been released as of the source report’s October 20 publication. MiniMax said the weights would arrive on October 23, but the supplied material does not confirm whether that happened or identify a license.
Did M3.1 Flash win the newsroom test?
The available source does not provide the newsroom test scores. It describes a test of 24 articles, in English and German, assessed by two judges, but the supplied material cuts off before the results.
How did independent speed testing compare with MiniMax’s claim?
MiniMax claimed everyday speeds above 200 tokens per second and a median above 170. Artificial Analysis measured 120 tokens per second and 1.08 seconds to first token in its evaluation; these figures come from different testing contexts.
Source: ThorstenMeyerAI.com
Fall Picks
fall essentials
As an affiliate, we earn on qualifying purchases.
