🔍 Read the full analysis: The Leaderboard Rank And Newsroom Results For MiniMax M3.1 Flash on ThorstenMeyerAI.com
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. Artificial Analysis scored it 43 on its Intelligence Index, while the supplied newsroom test description gives the setup but no results table or scores, so its writing performance cannot be assessed from the available material.
MiniMax released M3.1 Flash on October 19, setting prices at $0.15 per million input tokens and $0.60 per million output tokens. The launch positions the model as a lower-cost option for agent tasks, but its performance picture is mixed: Artificial Analysis assigned it an Intelligence Index score of 43, behind several open-weight competitors, and the source describing a 24-article newsroom test does not provide the test’s scores.
The model is available through the MiniMax API, M Plan subscription, MiniMax Code and OpenRouter. OpenRouter offered access without charge for the first 72 hours, according to the source. The listed price for cached input is $0.015 per million tokens. OpenRouter lists the predecessor M3 at $0.30 per million input tokens and $1.20 per million output tokens, making M3.1 Flash half the price per token on those measures.
MiniMax says the model reaches more than 200 tokens per second in everyday use, with a median above 170, and describes it as nearly three times faster than M3. Artificial Analysis measured a different result: 120 tokens per second and 1.08 seconds to first token. Those figures come from different measurement sources and should not be treated as directly interchangeable. The model has a one-million-token context window; OpenRouter lists a maximum output of 524,288 tokens.
The company says the model uses “more aggressive sparsity” but has not published an official parameter count or released the weights. MiniMax says weights are due October 23. Until then, Artificial Analysis lists M3.1 Flash as proprietary. The source also reports that the model spent ten days under the name Space Bunny Alpha on OpenRouter, where it led the platform’s usage ranking and processed 56.7 trillion tokens before being removed on October 5.
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 Uneven Benchmark Results
The launch matters to developers weighing cost, speed and task quality. The token price is half that of M3, but a lower unit price does not by itself establish lower total costs: Artificial Analysis says M3.1 Flash used 180 million tokens to complete its index evaluation, compared with a median of 99 million. That result suggests token consumption is another factor to examine alongside the advertised price.
On Artificial Analysis’s index, M3.1 Flash scored 43, compared with 45 for DeepSeek V4.1 Flash 1017 and GLM-5.3, and 46 for Xiaomi’s MiMo-V2.6-Pro. Its measured speed of 120 tokens per second was also below DeepSeek’s reported 211. These figures do not establish how the models will perform on every workflow, but they place MiniMax’s product behind several listed rivals on this particular evaluation.
For newsroom and publishing teams, the separate test could address a different question: whether the model can turn source material into accurate, useful copy without adding unsupported details. The described evaluation used real articles and two judges, but no scores or model comparisons are supplied. It cannot support a conclusion about M3.1 Flash’s newsroom quality as presented here.
As an affiliate, we earn on qualifying purchases.
From Space Bunny to M3.1 Flash
MiniMax’s October 19 announcement identified M3.1 Flash as an early version of the model that had appeared anonymously on OpenRouter as Space Bunny Alpha. According to the source, it went live under that name on September 23, led OpenRouter’s usage ranking for ten days and processed 56.7 trillion tokens there before its removal on October 5. MiniMax says the model processed more than 200 trillion tokens overall; that figure is the company’s account, not an independently verified measure in the supplied material.
MiniMax’s published benchmark chart compared M3.1 Flash with selected models, including DeepSeek V4.1 Flash 1017, Kimi K3, GLM-5.3, GPT-5.6 Sol and Claude Opus 5. The source notes that some newer versions were not included. In the reported results, M3.1 Flash scored 62.2 on Automation-Bench, ahead of DeepSeek’s 60.5, while DeepSeek led on DeepSWE and Terminal-Bench. The figures were transcribed from MiniMax chart images and are company-reported; the source says they have not been independently reproduced.
MiniMax also claimed first place among Chinese models on Agents’ Last Exam. The source cautions that MiniMax’s chart did not include DeepSeek in that row, while DeepSeek’s model card reports 35.4 on the benchmark, potentially under a different setup. That makes the ranking claim difficult to compare directly. The supplied material also says MiniMax’s security results, including an 82% CyberGym score, came from an internal evaluation.
““more aggressive sparsity””
— MiniMax
Weights, Data and Test Scores Pending
Several points remain unresolved. The weights have not yet been released, and the supplied material gives no confirmed license or official parameter count. A reported estimate of 428 billion total parameters and 23 billion active parameters is attributed to another outlet, but MiniMax has not confirmed it, so it cannot be treated as an established specification.
The source raises a data-handling question about the anonymous OpenRouter period: it says the provider could retain prompts but could not use them for training, and that MiniMax has not said what happens to that retained data. The available material also does not include the newsroom benchmark’s results table, scores or ranking. It describes 24 articles, two language versions and two judges, but does not establish how M3.1 Flash performed against other models in that test.
Finally, MiniMax’s benchmark results and speed claims are not equivalent to independent reproduction. The comparisons use selected rivals and, according to the source, some are older model versions. Differences in setup may also affect comparisons such as the Agents’ Last Exam result. These limits leave the model’s standing dependent on the benchmark and measurement method.
The October 23 Weights Release
The next stated milestone is October 23, when MiniMax says it plans to release the weights. The release would allow users to inspect the license and published model details, if provided, and could make it possible for outside evaluators to test the model in additional settings. The source does not confirm that the schedule is guaranteed or specify what license will apply.
Readers comparing M3.1 Flash with alternatives can watch for independent evaluations using documented, comparable methods, as well as clarification on prompt retention during the Space Bunny Alpha period. A complete newsroom results table is also needed before the described test can support conclusions about factual accuracy, writing quality or relative performance. For now, the confirmed record is a low per-token launch price, a mixed set of benchmark comparisons and open questions about weights, data handling and the missing newsroom results.
Key Questions
When did MiniMax release M3.1 Flash?
MiniMax released M3.1 Flash on October 19, 2026, according to the supplied source. It said the model’s weights would arrive on October 23.
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.
How did M3.1 Flash score on the independent index?
Artificial Analysis gave it an Intelligence Index score of 43. The source lists several competitors at 44 to 46, including DeepSeek V4.1 Flash 1017 at 45.
What did the newsroom test find?
The source describes a test of 24 articles: 12 real news sources written in English and German, with two judges scoring each model. However, it supplies no result table or scores for M3.1 Flash, so the test’s outcome cannot be reported from the available material.
Are the model weights available?
Not according to the source dated October 20. MiniMax said the weights would be released on October 23, but the available material does not confirm that they have since been published or specify a license.
Source: ThorstenMeyerAI.com
