🔍 Read the full analysis: A Leaderboard Lags Behind Our Newsroom Test Of MiniMax M3.1 Flash 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, 2026, at $0.15 per million input tokens and $0.60 per million output tokens. The source review says independent index results place it behind several open-weight competitors, while its newsroom test covers 24 articles; the supplied material does not include that test’s scores. The model’s weights are promised for October 23, and its speed and benchmark claims have not been independently reproduced in full.
MiniMax released M3.1 Flash on October 19, pricing it at $0.15 per million input tokens and $0.60 per million output tokens. A review by ThorstenMeyerAI.com compared the model’s company-reported benchmarks with Artificial Analysis measurements and a 24-article newsroom test, finding that the available results do not support treating MiniMax’s speed and performance claims as settled.
MiniMax describes M3.1 Flash as a fast, low-cost model for agent work. The company says it runs at more than 200 tokens per second in everyday use, with a median above 170, and is nearly three times faster than its predecessor, M3. The model has a one-million-token context window; MiniMax has not yet released its weights or confirmed a parameter count. It says the weights will arrive on October 23.
The review reports that Artificial Analysis gave the model an Intelligence Index score of 43, below several listed open-weight models, including Xiaomi’s MiMo-V2.6-Pro at 46 and DeepSeek V4.1 Flash 1017 and GLM-5.3 at 45. Artificial Analysis measured M3.1 Flash at 120 tokens per second, with 1.08 seconds to first token. Those results are lower than MiniMax’s speed claims, although the measures and test conditions are not necessarily identical.
The newsroom evaluation used 12 real news sources in English and German, producing 24 articles. Two judges assessed source fidelity, invented facts, clarity, usefulness, structure and polish. However, the supplied source material ends before giving the test’s scores or comparisons with other models. It therefore supports describing the test setup, but not claiming that M3.1 Flash passed or failed it.
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
The launch puts a low per-token price beside a practical question for teams choosing models: how much reliable work does each token buy? Artificial Analysis found the model used 180 million tokens to complete its index, compared with a median of 99 million. That is a measurement from one evaluation, not a general finding about every task, but it suggests that a low token price alone may not capture total operating cost.
For newsrooms and other users working from source documents, factual fidelity matters alongside speed and cost. The described newsroom test is relevant because it asks judges to compare generated stories directly with source material. Without the reported scores, readers cannot use it to rank M3.1 Flash against alternatives. The available evidence instead shows why vendor benchmarks, independent indexes and task-specific tests should be read as separate forms of evidence.
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 announcement, the model appeared anonymously on OpenRouter as Space Bunny Alpha. According to the review, it went live 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 it processed more than 200 trillion tokens overall. These figures describe platform usage, not a measure of model quality.
MiniMax also says M3.1 Flash costs half as much per token as M3 and about 70% less per agent task. Those are company-reported comparisons; the supplied material does not provide an independently verified task-cost study. MiniMax’s published benchmark charts, as transcribed in the review, show M3.1 Flash ahead on Automation-Bench, while DeepSeek V4.1 Flash leads on DeepSWE and Terminal-Bench. The company selected the competitors, and some newer rival versions were not included.
“More than 200 tokens per second in everyday use, with a median above 170.”
— MiniMax
As an affiliate, we earn on qualifying purchases.
Test Scores and Data Handling
The newsroom test’s English and German scores are not present in the supplied material, nor are the model-by-model results needed to judge its relative performance. The review says it used GPT-6.1 Sol and Claude Opus 5.5 as judges, but no scoring outcomes or disagreement between judges are provided here.
Other details also remain unsettled. MiniMax has not confirmed the model’s parameter count, and the source says its weights and license were not yet available. Artificial Analysis’s index and speed figures are independent measurements, but they reflect that organization’s evaluation methods. The review also notes that prompts sent to Space Bunny Alpha could be retained during its anonymous OpenRouter period, although the provider was not allowed to train on them; MiniMax has not clarified what happened to that data.
As an affiliate, we earn on qualifying purchases.
Weights and Results Due
The next stated milestone is October 23, when MiniMax says it will release the weights. Their availability, license and any confirmed model specifications will help establish how the release can be inspected and deployed. The newsroom review’s missing scores are also needed before readers can assess its central practical comparison. Until those details are available, the clearest picture is a low-priced release with mixed independent indicators and a company speed claim that exceeds the reported index measurement.
real news article AI evaluation tools
As an affiliate, we earn on qualifying purchases.
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 subscription, MiniMax Code and OpenRouter, according to the source review.
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.
How did it score on the newsroom test?
The supplied material describes a test of 24 articles based on 12 sources, in English and German, judged by GPT-6.1 Sol and Claude Opus 5.5. It does not include the scores, so the outcome cannot be established from this source.
Are M3.1 Flash’s weights available?
Not as of the October 20 review. MiniMax said the weights would arrive October 23; the source does not confirm that release or provide a license.
Does the independent index confirm MiniMax’s speed claim?
No. MiniMax claimed more than 200 tokens per second in everyday use, while Artificial Analysis measured 120 tokens per second in its evaluation. The figures come from different testing contexts and do not, by themselves, establish performance across all workloads.
Source: ThorstenMeyerAI.com
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
