MiniMax M3.1 Flash In The Newsroom: Results Beyond The Leaderboard

🔍 Read the full analysis: MiniMax M3.1 Flash In The Newsroom: Results Beyond The Leaderboard on ThorstenMeyerAI.com

Before you orderOffer from Amazon

Get smart everyday buys delivered free with Prime

• Fast, free delivery on millions of items
• Prime Video, Amazon Music and more included
• Member-only deals all year

Start your free Prime trial →Free trial for eligible customers · Cancel anytime

As an affiliate, we earn on qualifying purchases.Hide

MiniMax released M3.1 Flash on October 19, offering the model through its API, M Plan subscription, MiniMax Code and OpenRouter at $0.15 per million input tokens and $0.60 per million output tokens. The release puts MiniMax’s claims of speed and agent efficiency against independent measurements that show a capable but not leading model—and a newsroom test whose results are not included in the supplied material.

MiniMax describes M3.1 Flash as a low-cost model for agent tasks. The company says it is nearly three times faster than its predecessor, M3, in everyday use, with speeds above 200 tokens per second and a median above 170. It also says the model costs about 70% less per agent task than M3. Those are company claims; an independent evaluation reported a lower output speed.

The listed prices are $0.15 per million input tokens, $0.60 per million output tokens and $0.015 per million cached input tokens. OpenRouter initially made the model available free for 72 hours. MiniMax lists a one-million-token context window; OpenRouter permits up to 524,288 output tokens, and users can adjust reasoning effort.

MiniMax has not released the weights yet. It says they will arrive on October 23. The company has not published a license or confirmed a parameter count. A separate outlet reportedly gave a parameter estimate, but the source review says MiniMax has not confirmed it, so that figure cannot be treated as established.

Price Meets Real-World Performance

The release matters to teams weighing model cost against task quality, especially for agent workflows that can consume many tokens. M3.1 Flash’s listed token prices are half M3’s, according to the comparison in the source review. But price per token does not by itself establish lower total costs: the review reports that the model used 180 million tokens to complete Artificial Analysis’s index, compared with a 99-million-token median. That is a raw comparison of token use, not a measure of equivalent task cost.

For newsroom use, the key question is whether the model can turn source material into accurate copy without adding unsupported details. The test described in the source used 24 articles in English and German, judged by two separate models. However, the provided material ends before giving the scores. It therefore supports describing the test design, but not claiming that M3.1 Flash passed, failed or ranked against other models.

The results also caution against treating a leaderboard position or a low listed price as a complete measure of usefulness. Independent speed and intelligence scores, token consumption, accuracy on real assignments and the eventual weight license all affect how a newsroom or developer might assess the model.

From Space Bunny to M3.1

Before the announcement, the model appeared anonymously on OpenRouter as Space Bunny Alpha. According to the source review, it was available from September 23, led OpenRouter’s usage ranking for 10 days and processed 56.7 trillion tokens on the platform before being removed on October 5. MiniMax says total use across services exceeded 200 trillion tokens. These figures are attributed to the review and MiniMax, respectively, and refer to different scopes.

The anonymous deployment has a privacy dimension. The review says the provider could retain prompts during that period but was not allowed to use them for training. MiniMax has not explained what happened to the retained data. The source also reports that MiniMax’s published benchmark comparisons rely on company-selected rivals and self-reported scores. In several comparisons, newer versions of rival models were not included; the review says the charts do not report every rival on every test.

Artificial Analysis, which evaluates models using the same set of ten assessments, gave M3.1 Flash an Intelligence Index score of 43. The source lists higher scores for several open-weight models, including DeepSeek V4.1 Flash 1017 at 45 and MiMo-V2.6-Pro at 46. Artificial Analysis measured M3.1 Flash at 120 tokens per second and 1.08 seconds to first token—below MiniMax’s stated everyday speed. The index classified it as proprietary because its weights had not yet been released.

“The weights will arrive on Oct. 23.”

— MiniMax

Weights, Data and Test Scores

Several points remain open. The promised October 23 weight release had not happened as of the October 20 source date, and MiniMax had not stated a license or official parameter count. Without those details, developers cannot yet assess the exact terms for using or modifying the weights.

The source review says MiniMax has not clarified what happened to prompts retained during the Space Bunny Alpha period. It also says the company’s security results, including an 82% score on CyberGym, came from an internal evaluation and had not been independently reproduced. Those results should be understood as company-reported, not independently verified.

The source material describes the newsroom benchmark’s 24-article setup and names the two judges, but its results table is cut off before any scores appear. It is therefore unclear how M3.1 Flash performed on source fidelity, fabricated details, clarity, usefulness, structure or polish, and no comparison with other models can be reported from the supplied information.

The October 23 Weight Release

The next stated milestone is MiniMax’s planned release of M3.1 Flash’s weights on October 23. Whether they arrive on schedule—and what license accompanies them—will affect whether outside researchers and developers can inspect, reproduce or adapt the model.

Further evaluation could clarify how the model performs beyond company-selected benchmarks. That includes independent replication of reported security results, comparisons against current rival versions under matching conditions, and publication of the missing newsroom-test scores. Until then, the clearest evidence is the mix already available: low listed prices, a 43-point independent index score, and a measured speed below MiniMax’s stated figure.

Key Questions

When did MiniMax release M3.1 Flash?

MiniMax announced and released M3.1 Flash on October 19, 2026, through its API, M Plan subscription, MiniMax Code and OpenRouter.

How much does M3.1 Flash cost?

The listed rate is $0.15 per million input tokens, $0.60 per million output tokens and $0.015 per million cached input tokens. OpenRouter offered it free for its first 72 hours.

Are M3.1 Flash’s weights available?

Not according to the October 20 source. MiniMax said the weights would arrive October 23, but the company had not yet provided them or announced a license in the material reviewed.

How fast is the model?

MiniMax claims 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, so reported results depend on the measurement method.

What did the newsroom test find?

The test covered 24 articles: 12 sources written in English and German, then assessed by GPT-6.1 Sol and Claude Opus 5.5. The supplied material does not include the results table, so it does not establish how the model scored.

Source: ThorstenMeyerAI.com