🔍 Read the full analysis: A Newsroom Test Of MiniMax M3.1 Flash Versus Open Models 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, at $0.15 per million input tokens and $0.60 per million output tokens, with model weights promised for October 23. Artificial Analysis measured an Intelligence Index score of 43 and output speed of 120 tokens per second, while the supplied newsroom test description gives its methodology but no results table. The evidence supports a lower listed price, but does not yet establish how M3.1 Flash performs against other models in real newsroom use.
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 lower-cost model aimed at agent tasks. A report by ThorstenMeyerAI.com compared the company’s claims with Artificial Analysis measurements and a test using 24 news articles, but the supplied material does not include the newsroom test’s scores.
The release followed a period when a model identified as Space Bunny Alpha appeared anonymously on OpenRouter. The report says it led the platform’s usage ranking for 10 days, from September 23, and processed 56.7 trillion tokens on OpenRouter before being removed on October 5. MiniMax later identified it as an early version of M3.1 Flash. The company says the model processed more than 200 trillion tokens overall.
M3.1 Flash is available through the MiniMax API, M Plan subscription, MiniMax Code and OpenRouter. OpenRouter offered it without charge for its first 72 hours. The listed price is half the per-token price of M3, according to the report; MiniMax also claims the new model costs about 70% less per agent task. The latter is a company claim, not an independently verified cost comparison.
MiniMax advertises a one-million-token context window and says everyday output speed exceeds 200 tokens per second, with a median above 170. Artificial Analysis measured 120 tokens per second and a 1.08-second time to first token. The organization gave M3.1 Flash an Intelligence Index score of 43, below the listed scores of several open-weight models. Its evaluation also found the model used 180 million tokens to complete the index, compared with a 99-million-token median.
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 Versus Measured Performance
The release adds another low-priced option for developers building AI agents and other token-intensive workflows. But a low charge per token does not by itself establish a lower total cost: a model that uses more tokens to complete a task can reduce or erase the apparent savings. Artificial Analysis’s finding that M3.1 Flash used 180 million tokens on its index, compared with a 99-million-token median, highlights that distinction, though it is not a direct measure of every real-world task.
The performance gap also matters to teams choosing models for speed-sensitive work. MiniMax’s advertised speed is above 200 tokens per second, while Artificial Analysis recorded 120. The measurement methods and conditions may differ, and the available report does not reconcile them. For now, buyers have a price and some comparative evaluation data, but not a full, independently verified account of what the model will cost or how quickly it will run in their own applications.
The promised weights could change the model’s status and make outside inspection possible. Until they are released, M3.1 Flash is listed by Artificial Analysis as proprietary. MiniMax has not provided a license or an official parameter count in the source material.
As an affiliate, we earn on qualifying purchases.
From Space Bunny to M3.1
The anonymous OpenRouter listing drew attention before MiniMax named the model. According to the report, Space Bunny Alpha topped OpenRouter’s usage ranking for 10 days and was taken down on October 5. MiniMax identified it publicly as an early version of M3.1 Flash on October 19. The timing links the anonymous model’s platform use to the new release, but does not establish that every detail of the earlier deployment matches the final product.
MiniMax’s own benchmark charts compare M3.1 Flash with a selected group of models. In those results, DeepSeek V4.1 Flash 1017 scored higher on DeepSWE and Terminal-Bench, while M3.1 Flash scored 62.2 on Automation-Bench, ahead of DeepSeek’s 60.5. The report notes that MiniMax’s chart did not include DeepSeek on its Agents’ Last Exam row, where MiniMax describes its result as first among Chinese models. The scores are company-reported, and the supplied material says they have not been independently reproduced.
ThorstenMeyerAI.com also describes a newsroom test: 12 source articles were each written in English and German, producing 24 outputs. Two judges, GPT-6.1 Sol and Claude Opus 5.5, scored each output from 1 to 5 on source fidelity, invented facts, clarity, usefulness, structure and polish. The stated setup used the production prompt and a free OpenRouter endpoint with a 32,000-token output limit. The source excerpt ends before supplying the results table, so it does not support a conclusion about how M3.1 Flash scored or ranked.
““more aggressive sparsity””
— MiniMax
As an affiliate, we earn on qualifying purchases.
Missing Scores and Data Details
The supplied account does not include the newsroom benchmark’s score table. It therefore remains unclear how M3.1 Flash performed on the 24 articles, how it compared with open-weight models and closed references, or whether the two judges agreed. The methodology alone cannot establish the model’s newsroom accuracy or reliability.
Other open questions include the conditions behind MiniMax’s speed claims and how they compare with Artificial Analysis’s measurements. MiniMax has not disclosed an official parameter count or weight license in the source material, and the promised weights were not yet released as of the report’s October 20 publication date. The report also says that during the anonymous OpenRouter period, the provider could retain prompts but could not use them for training; MiniMax has not stated what happened to that retained data.
Comparisons across company benchmark charts and independent evaluations should be treated cautiously. The source says MiniMax selected its benchmark rivals, omitted some newer competitors, and reported figures that had not been independently reproduced. Different test setups may also affect results.
As an affiliate, we earn on qualifying purchases.
Weights Due October 23
The next stated milestone is MiniMax’s planned October 23 release of M3.1 Flash’s weights. Their publication could provide details now missing from the report, including a license and information that may help researchers inspect the model. The material does not confirm that the release has occurred.
For the newsroom comparison, the next useful evidence would be the omitted results table, including scores by language, model and judging criterion. That would allow readers to assess source fidelity and fabrication risk rather than infer performance from the test design. Independent replication of MiniMax’s benchmark and security claims would also help clarify how its results compare with other models.
Until those details are available, the confirmed case for M3.1 Flash is its release and its lower listed token price than M3. Its broader performance, total task cost, open-weight status and handling of prompts from the anonymous period remain subjects for verification.
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 report. The company 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 and $0.60 per million output tokens, plus $0.015 per million cached input tokens. The report lists M3 at $0.30 per million input tokens and $1.20 per million output tokens.
Did the newsroom test show that M3.1 Flash is accurate?
The source describes a 24-article test judged on source fidelity, invented facts, clarity, usefulness, structure and polish, but does not provide the results table. No conclusion about the model’s score or ranking can be drawn from the supplied information.
Is M3.1 Flash open-weight?
Not yet, according to the report. The model was listed as proprietary by Artificial Analysis because its weights had not been released. MiniMax said the weights were due on October 23; the material does not confirm their later release.
How fast is M3.1 Flash?
MiniMax claims everyday speed above 200 tokens per second and a median above 170. Artificial Analysis measured 120 tokens per second. The supplied material does not establish whether those figures came from comparable testing conditions.
Source: ThorstenMeyerAI.com
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
