🔍 Read the full analysis: MiniMax M3.1 Flash: The Newsroom Test Behind Its Open-Model Results on ThorstenMeyerAI.com
Get business pricing on office and shipping supplies
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
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 source’s benchmark review reports mixed results: MiniMax’s own tests show strengths on one automation benchmark, while Artificial Analysis scores it below several open-weight rivals. A newsroom test covered 24 articles, but the supplied source ends before giving its scores, so its results cannot be reported.
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 review by ThorstenMeyerAI.com compared MiniMax’s claims with Artificial Analysis measurements and a newsroom test of 24 articles; the results described in the source show a capable but not clearly leading model, while the newsroom scores themselves are not included in the supplied material.
MiniMax says M3.1 Flash is designed for fast, lower-cost agent work and costs half as much per token as its predecessor, M3. The company also says it costs about 70% less per agent task. Those figures are MiniMax’s claims. It has not published technical details such as an official parameter count, and its model weights were still pending as of the October 20 review.
On Artificial Analysis’s Intelligence Index, M3.1 Flash scored 43. That trailed the listed scores for MiMo-V2.6-Pro at 46, DeepSeek V4.1 Flash 1017 and GLM-5.3 at 45, and Step 5 Preview and Kimi K3 at 44. Artificial Analysis measured its output speed at 120 tokens per second, below MiniMax’s claim of more than 200 tokens per second in everyday use. The review also reports that M3.1 Flash used 180 million tokens to complete the index, compared with a 99 million-token median.
MiniMax’s own charts show a more mixed picture. M3.1 Flash scored 62.2 on Automation-Bench, ahead of the listed competitors, but DeepSeek V4.1 Flash 1017 led it on DeepSWE and Terminal-Bench. The review says MiniMax selected the models in its comparison and notes that some newer rival versions were not included. The newsroom test used 12 news sources in English and German, producing 24 articles scored by two judges, but the supplied source cuts off before reporting any model results from that test.
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, Speed and Real-World Accuracy
M3.1 Flash is priced for lower-cost agent use, but per-token pricing does not establish the cost of completing a task. Artificial Analysis found that the model used more tokens than the index median, a factor that could affect total costs. The review does not establish typical user costs across different workloads.
The newsroom test results are not included in the supplied material. The test was designed to assess how models turn source material into written output, using real news sources in two languages and two judges. Its scores could provide evidence about source fidelity and fabricated details, but no results are available in the source.
The release was not yet an open-weight release as of the October 20 review. MiniMax said the weights were planned for October 23, but had not announced a license or confirmed the model’s parameter count. Those details are needed to assess the terms of access and whether users can independently inspect the weights.
As an affiliate, we earn on qualifying purchases.
From Space Bunny to M3.1 Flash
Before its public launch, the model appeared anonymously on OpenRouter as Space Bunny Alpha. According to the review, 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 later said the model had processed more than 200 trillion tokens overall; that total is the company’s figure.
The review says OpenRouter allowed the provider to retain prompts during the anonymous period, though not to use them for training. MiniMax had not explained what happened to that data. On release, M3.1 Flash became available through the MiniMax API, M Plan subscription, MiniMax Code and OpenRouter. OpenRouter offered it free for its first 72 hours, according to the source.
MiniMax claims everyday output speeds above 200 tokens per second and a median above 170, describing the model as nearly three times faster than M3. Artificial Analysis’s measured speed was 120 tokens per second, a different result from the company’s claim. The review notes that MiniMax attributes the design to “more aggressive sparsity” but provides no further technical explanation.
““more aggressive sparsity””
— MiniMax
Missing Scores and Unreleased Weights
The supplied source does not include the newsroom benchmark’s results: it ends after introducing a table of models, without giving scores. No conclusion about M3.1 Flash’s performance on the 24 articles can be drawn from this material. The described test used two judges, GPT-6.1 Sol and Claude Opus 5.5, who scored source fidelity, invented facts, clarity, usefulness, structure and polish, but the individual ratings and comparisons are missing.
Other points remain unresolved. MiniMax had not confirmed a parameter count, published a license or explained what happened to prompts retained during the model’s anonymous OpenRouter period. Its benchmark charts are company-selected and self-reported, and the review says the security results had not been independently reproduced. The company’s speed and cost claims should also be distinguished from Artificial Analysis’s measurements, which use a separate evaluation.
October 23 Weights Release
The next stated milestone is October 23, 2026, when MiniMax says it plans to release M3.1 Flash’s weights. The release would allow users to examine the model under the terms of its eventual license; neither the license nor the exact release status was confirmed in the October 20 source.
Further reporting could compare the newsroom scores once they are available, verify the company’s benchmark claims independently and clarify how the model handled prompts during its anonymous deployment. Until then, the available information establishes that M3.1 Flash is a low-priced API model with mixed benchmark results; its open-weight status and newsroom performance remain unconfirmed.
Key Questions
When did MiniMax release M3.1 Flash?
MiniMax released M3.1 Flash on October 19, 2026, according to the October 20 review. The model was available through the MiniMax API, M Plan, 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, according to the source.
Is M3.1 Flash open-weight?
Not yet, based on the source. MiniMax said the weights would arrive on October 23; as of the October 20 review, no license or official parameter count had been published.
How did M3.1 Flash perform against other models?
Artificial Analysis gave it an Intelligence Index score of 43, below several listed open-weight rivals, and measured output speed at 120 tokens per second. MiniMax’s own charts showed it leading Automation-Bench, while DeepSeek V4.1 Flash 1017 led on DeepSWE and Terminal-Bench.
What did the newsroom test find?
The review describes a test of 24 articles based on 12 news sources, with English and German versions scored by two judges. The supplied source does not include the scores, so the test’s findings cannot be stated.
Source: ThorstenMeyerAI.com
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
