🔍 Read the full analysis: MiniMax M3.1 Flash: How It Performed In A Real Newsroom Test on ThorstenMeyerAI.com
Prime made for students and young adults
- Fast, free delivery for dorm and study essentials
- Prime Video and Amazon Music included
- Member-only deals
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. ThorstenMeyerAI.com tested it on 24 news articles, but the supplied source cuts off before reporting the newsroom scores, so its performance in that test cannot be established from the available material. An independent index gave it a score of 43, behind several open-weight competitors, while MiniMax’s speed and benchmark claims have not been independently reproduced in the source.
MiniMax lists M3.1 Flash at **$0.15 per million input tokens** and **$0.60 per million output tokens**, with cached input priced at $0.015 per million tokens. Those rates are half the listed per-token price of M3, according to the source. The company also says the new model costs about **70% less per agent task** than its predecessor, a claim that depends on task-level usage and is not independently verified in the supplied material.
MiniMax describes the model as a fast, inexpensive option for agent work and claims everyday generation 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 Intelligence Index score was **43**, compared with 45 for DeepSeek V4.1 Flash 1017 and GLM-5.3, and 46 for Xiaomi’s MiMo-V2.6-Pro.
For its newsroom evaluation, ThorstenMeyerAI.com used **12 real news sources**, ranging from a short press release to a detailed report. Each source was turned into an English and a German article using the publication’s production prompt, for **24 articles total**. Two judges, GPT-6.1 Sol and Claude Opus 5.5, scored articles from 1 to 5 for source fidelity, invented facts, clarity, usefulness, structure and polish. The provided source ends before its results table, so it does not establish the model’s scores or how it compared with other tested models.
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
The Gap Between Price and Proof
The release puts a **low-priced model aimed at agent tasks** into a field where cost per token is only part of the practical calculation. Artificial Analysis reported that M3.1 Flash used **180 million tokens** to complete its index evaluations, against a median of 99 million. That observation suggests token volume could affect total cost, but it does not by itself establish the model’s cost on a newsroom assignment or other real-world task.
For publishers, the unanswered newsroom scores matter more than a general benchmark claim: a model must preserve source details and avoid invented facts to be useful in reporting workflows. The test design addresses those concerns, but without its results readers cannot judge whether M3.1 Flash met them. The discrepancy between MiniMax’s claimed speed and Artificial Analysis’s measured speed also makes independent, task-specific comparisons relevant.
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 the public release, the model appeared anonymously on OpenRouter as **Space Bunny Alpha** beginning September 23, 2026. The source says it led OpenRouter’s usage ranking for ten days, processed 56.7 trillion tokens there, and was removed on October 5. MiniMax later said the model had processed more than **200 trillion tokens overall**; that total is the company’s figure.
MiniMax says the model has a **one-million-token context window** and adjustable reasoning effort. The weights were not available at release: the company said they would arrive on **October 23**. Until weights are released, Artificial Analysis classifies M3.1 Flash as proprietary. No official parameter count or license was provided in the source.
The review also cautions that MiniMax’s published benchmark charts compare the model with selected competitors and rely on company-reported scores. For example, MiniMax reported a 62.2 on Automation-Bench, while DeepSeek V4.1 Flash 1017 led the listed coding results on DeepSWE and Terminal-Bench. These are not a substitute for the missing newsroom test outcome.
“The company said M3.1 Flash would deliver “more than 200 tokens per second” in everyday use, with a median above 170.”
— MiniMax
The Missing Newsroom Scores
The supplied source material stops after introducing the newsroom results table and does not include any scores, rankings or judge comments. It is therefore **not possible to report whether M3.1 Flash performed well** on source fidelity, factual accuracy or writing quality, or how it compared with the open-weight and closed models in the test.
Other points remain unsettled. MiniMax’s speed and agent-cost figures are company claims, and the source says its benchmark and security results had not been independently reproduced. MiniMax had not confirmed a parameter count or license, and the review says it had not explained what happened to prompts retained during the model’s anonymous OpenRouter period. The source states those prompts could be kept but not used for training.
Weights and Full Test Results
The next stated milestone was MiniMax’s planned **October 23 weights release**. Whether that release occurred, what license accompanied it, and whether outside researchers reproduced the model’s results are not established by the October 20 source.
The newsroom comparison also needs its missing results table before readers can assess the model against the other systems tested. A complete account should report the English and German scores, explain how the judges’ ratings were combined, and identify any differences between the model outputs. Until then, the test method is documented, but its central performance finding remains unavailable.
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 subscription, MiniMax Code and OpenRouter.
What 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. MiniMax separately claims about 70% lower cost per agent task than M3.
How did it perform in the newsroom test?
The review describes a test of **24 articles** in English and German, judged for source fidelity, invented facts, clarity, usefulness, structure and polish. The supplied source omits the results, so its newsroom performance cannot be determined from the available information.
Is M3.1 Flash open-weight?
Not at the time covered by the source. MiniMax said weights would arrive on **October 23, 2026**, and no license was given in the material. The source does not confirm whether the planned release later occurred.
Source: ThorstenMeyerAI.com
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
