🔍 Read the full analysis: MiniMax M3.1 Flash: A Newsroom Test Tells A Different Story Than The Leaderboard on ThorstenMeyerAI.com
Get hardware and tech essentials 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, promoting it as a fast, low-cost model for agent tasks. An independent index gave it a score of 43, below several open-weight rivals, while a newsroom test covered 24 articles; the supplied source does not include that test’s scores or results. The model’s weights are scheduled for October 23, and its reported speed, benchmark claims and handling of prompts from its anonymous OpenRouter period remain points to verify.
MiniMax released M3.1 Flash on October 19 at a lower per-token price than its predecessor, M3, while marketing it for agent work. A review published by ThorstenMeyerAI.com compared MiniMax’s claims with Artificial Analysis measurements and a 24-article newsroom test; the available source describes the test setup but does not provide its scores or conclusions, so it cannot establish how the model performed against other systems on that task.
MiniMax lists M3.1 Flash at $0.15 per million input tokens and $0.60 per million output tokens, with cached input priced at $0.015 per million tokens. Those rates are half the listed input and output prices for M3. The model is available through the MiniMax API, M Plan subscription, MiniMax Code and OpenRouter; OpenRouter offered it free for its first 72 hours. MiniMax says the new model costs about 70% less per agent task, but that figure is a company claim, not an independently verified result in the source.
MiniMax says everyday output speed exceeds 200 tokens per second, with a median above 170, and describes the model as nearly three times faster than M3. Artificial Analysis measured 120 tokens per second and 1.08 seconds to first token in its evaluation. The same index assigned M3.1 Flash an Intelligence Index score of 43, below DeepSeek V4.1 Flash 1017 and GLM-5.3 at 45, as well as several other listed open-weight models. The measurements come from different testing setups, so the speed figures should not be treated as a direct, controlled comparison.
The newsroom review says it tested 12 real news sources in both English and German, producing 24 articles, with GPT-6.1 Sol and Claude Opus 5.5 scoring each article against its source. Judges rated fidelity, invented facts, clarity, usefulness, structure and polish. However, the supplied source cuts off before the result table and gives no newsroom scores or comparative findings. The article’s headline frames the test as differing from leaderboard results, but the missing data means that conclusion cannot be independently checked from the material provided.
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
Why newsroom accuracy matters
For publishers, model quality is not captured by speed or benchmark rank alone. A system used to draft news must preserve what a source says, avoid adding unsupported details and produce clear copy. The newsroom test’s focus on source fidelity and invented facts addresses those practical risks, but without its scores readers cannot tell whether M3.1 Flash met the test’s standard or how it compared with alternatives.
The price and speed claims also matter to teams running large volumes of agent or writing tasks. A lower token price may reduce costs, but Artificial Analysis reported that the model used 180 million tokens to complete its index, compared with a median of 99 million. That is a raw comparison reported for that evaluation, not proof that the model will use the same amount in other workloads. Buyers need task-specific measurements of quality, latency and total token use before drawing conclusions about cost.
As an affiliate, we earn on qualifying purchases.
From anonymous OpenRouter model to release
Before the public release, the model reportedly operated anonymously on OpenRouter as Space Bunny Alpha. The review says it appeared on September 23, led OpenRouter’s usage ranking for 10 days and processed 56.7 trillion tokens on the platform before being pulled on October 5. MiniMax says more than 200 trillion tokens were processed overall. These figures are reported in the review and attributed to the relevant platform or company; the source does not provide underlying records for independent verification.
The review also says that during the anonymous period the provider could retain prompts but was not allowed to train on them. MiniMax has not said what happened to that retained data, according to the source. The company has announced a one-million-token context window, adjustable reasoning effort and a planned release of weights on October 23. Until the weights are available, Artificial Analysis lists the model as proprietary. MiniMax has not disclosed an official parameter count or license.
““nearly three times faster” than M3”
— MiniMax
As an affiliate, we earn on qualifying purchases.
Missing newsroom scores and data details
The key gap is the newsroom benchmark’s outcome: the source provides its method but not the score table, model comparisons or judges’ findings. It is therefore not possible to verify what the headline calls a different story from the leaderboard. The source also does not establish whether MiniMax’s speed and cost claims will hold across different prompts, workloads or serving conditions.
Other open questions include the model’s weight release and license, its official parameter count, and the fate of prompts retained during the Space Bunny Alpha period. MiniMax’s benchmark chart omits some rival results, and the review says its security figures, including an 82% CyberGym result, came from internal evaluation and had not been independently reproduced. Those claims should not be read as independently confirmed performance.
As an affiliate, we earn on qualifying purchases.
Weights and full test results
MiniMax said the weights were due on October 23, 2026; their availability, license and technical details would clarify whether outside researchers can reproduce the reported results. The next useful evidence would also include the complete newsroom score table, judge-level results and comparisons with other models under the same prompts and conditions.
Until those details are available, readers can treat the release price and stated product features as published facts, while treating performance and task-cost figures as claims or measurements tied to particular evaluations. The source does not report a final independent audit of the newsroom test or a MiniMax explanation of the retained prompts.
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, according to the review. It was offered 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 and $0.60 per million output tokens. Cached input is listed at $0.015 per million tokens. These are token prices, not a guaranteed cost for completing a particular task.
Did the newsroom test show that M3.1 Flash was less reliable?
The supplied source describes a test of 24 articles but does not include the scores or findings. It is not possible to confirm from this material whether the model performed better or worse than other systems.
How did M3.1 Flash score on Artificial Analysis?
Artificial Analysis gave M3.1 Flash an Intelligence Index score of 43. The index listed DeepSeek V4.1 Flash 1017 and GLM-5.3 at 45. These are results from that index, not a direct measure of performance on every newsroom task.
When are the model weights expected?
MiniMax said the weights would arrive on October 23, 2026. The supplied source does not confirm whether they were released, or provide a license or official parameter count.
Source: ThorstenMeyerAI.com
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
