🔍 Read the full analysis: How MiniMax M3.1 Flash Went From Lower Ranking To Strong Test Results on ThorstenMeyerAI.com
Get business pricing on monitors, keyboards and dev gear
- 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, at half the per-token price of its predecessor and described it as a fast, low-cost model for agent tasks. Independent measurements put it behind several open-weight rivals on intelligence and speed, while the supplied source does not include the newsroom benchmark’s scores. Its weights are scheduled for October 23, but the license and parameter count have not been disclosed.
MiniMax released M3.1 Flash on October 19, pricing it at $0.15 per million input tokens and $0.60 per million output tokens as a model aimed at agent tasks. Early results present a mixed picture: Artificial Analysis measured a 43 Intelligence Index score and 120 tokens per second, while MiniMax’s own charts show wins on some tests and losses on others. The model’s weights are scheduled for release October 23, but have not yet been published.
MiniMax says M3.1 Flash costs half as much per token as M3 and about 70% less per agent task. The company claims everyday generation speeds above 200 tokens per second, with a median above 170, and describes the model as nearly three times faster than M3. Those speed and task-cost figures are company claims, not independent measurements. The model supports a one-million-token context window, adjustable reasoning effort and cached input priced at $0.015 per million tokens.
Artificial Analysis, which evaluates models on a shared set of ten tests, gave M3.1 Flash an Intelligence Index score of 43. That puts it below the open-weight models listed in the source, including DeepSeek V4.1 Flash 1017 and GLM-5.3, both at 45, and Xiaomi’s MiMo-V2.6-Pro at 46. Artificial Analysis measured M3.1 Flash at 120 tokens per second with 1.08 seconds to first token—below MiniMax’s stated speed and DeepSeek’s measured 211 tokens per second.
MiniMax’s published benchmark charts show a more varied result. M3.1 Flash scored 62.2 on Automation-Bench, ahead of the listed rivals, but DeepSeek V4.1 Flash 1017 scored higher on DeepSWE and Terminal-Bench. MiniMax’s scores are self-reported, and the source says they have not been independently reproduced. The supplied report also describes a newsroom test using 24 articles and two judges, but its results table is cut off; no scores or conclusions from that test can be reported here.
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 Gains Meet Performance Tradeoffs
M3.1 Flash’s low per-token price could make it attractive for developers running high-volume agent workflows. But the independent results show why price alone is an incomplete guide: Artificial Analysis found that the model used 180 million tokens to complete its index, compared with a 99-million-token median. The report does not establish that this difference will apply to every workload, but it indicates that total usage can affect the cost advantage.
For buyers, the useful comparison is between cost, speed and task quality on their own workload—not a single leaderboard score or a vendor’s headline speed claim. M3.1 Flash’s competitive Automation-Bench result may appeal to some users, while its lower index score than several listed open-weight rivals and its measured speed below DeepSeek’s may matter to others. Because the weights are still unavailable, users cannot yet inspect or deploy them as an open-weight model.
As an affiliate, we earn on qualifying purchases.
From Space Bunny to M3.1
Before its public identification, the model ran on OpenRouter under the name Space Bunny Alpha. The source says it appeared 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 the model processed more than 200 trillion tokens overall; those figures describe usage, not independent quality testing.
MiniMax identified Space Bunny Alpha as an early version of M3.1 Flash when it announced the model on October 19. During the anonymous period, the provider could retain prompts but was not allowed to train on them, according to the source. MiniMax has not clarified what happened to data retained during that period. The source also says OpenRouter offered M3.1 Flash free for its first 72 hours after release.
The release sits in a close price competition with DeepSeek V4.1 Flash 1017, which the source says appeared two days earlier at the same per-token price. MiniMax’s chart compares M3.1 Flash with selected models, but some comparison models are described as older generations. Scores from different test setups should not be treated as directly equivalent: for example, the report notes that DeepSeek’s score on Agents’ Last Exam comes from its model card, while MiniMax’s chart does not list DeepSeek for that row.
““nearly three times faster” than M3”
— MiniMax
high-performance AI inference server
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Weights, Data and Scores Await Clarity
The model weights, license and official parameter count are not yet available in the source. An outlet reportedly estimated 428 billion total parameters and 23 billion active parameters, but MiniMax has not confirmed those figures. Until the weights are released, M3.1 Flash is listed by Artificial Analysis as proprietary.
There is also no independent reproduction of the benchmark scores MiniMax published, according to the source. Its charts omit some rival scores, and comparisons may use different evaluation setups. The reported newsroom benchmark has no visible results in the provided material, so it is unclear how M3.1 Flash scored for factual accuracy, invented details, clarity or structure. MiniMax has not said what happened to prompts that could be retained during the model’s anonymous OpenRouter period.
As an affiliate, we earn on qualifying purchases.
October 23 Weights Release
The next stated milestone is October 23, 2026, when MiniMax says it plans to publish the M3.1 Flash weights. The release should clarify whether developers can access and run the model outside MiniMax’s services, and what license terms apply. MiniMax has not confirmed a parameter count in the source.
Further independent testing will be needed to establish how the model performs across repeatable benchmarks and real workloads. Readers should also look for the missing newsroom test results before drawing conclusions about its performance on source-based article writing. Until then, the confirmed picture is a low-cost release with mixed benchmark evidence and several important details still pending.
As an affiliate, we earn on qualifying purchases.
Key Questions
When did MiniMax release M3.1 Flash?
MiniMax announced and released M3.1 Flash on October 19, 2026, through the MiniMax API, M Plan subscription, MiniMax Code and OpenRouter, according to the source.
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. MiniMax says the per-token price is half that of M3.
Does M3.1 Flash outperform DeepSeek V4.1 Flash 1017?
The results are mixed. MiniMax’s charts show M3.1 Flash ahead on Automation-Bench, while DeepSeek scores higher on DeepSWE and Terminal-Bench. Artificial Analysis also measured DeepSeek at 45 on its Intelligence Index and 211 tokens per second, compared with M3.1 Flash at 43 and 120 tokens per second.
Are M3.1 Flash’s weights available?
Not in the source’s account of the release. MiniMax said the weights would arrive October 23. The source provides no license, and MiniMax has not confirmed an official parameter count.
What did the newsroom test find?
The source describes a test of 24 articles, assessed by two judges, but the results table is missing from the supplied material. Its scores and conclusions therefore cannot be confirmed from the information available here.
Source: ThorstenMeyerAI.com
Fall Picks
fall essentials
As an affiliate, we earn on qualifying purchases.
