Can A Newsroom Test Change How You See MiniMax M3.1 Flash?

🔍 Read the full analysis: Can A Newsroom Test Change How You See MiniMax M3.1 Flash? on ThorstenMeyerAI.com

MiniMax released M3.1 Flash on October 19, pricing the model at $0.15 per million input tokens and $0.60 per million output tokens. A review published October 20 compared MiniMax’s claims with Artificial Analysis measurements and a newsroom evaluation of 24 articles; the available report describes that test’s method but does not provide its scores or comparative findings.

MiniMax markets M3.1 Flash as a low-cost, fast model for agent work. Its listed price per token is half the stated M3 rate of $0.30 per million input tokens and $1.20 per million output tokens. The company also claims the new model is about 70% cheaper per agent task, a separate comparison that the report does not independently verify. MiniMax says everyday speeds exceed 200 tokens per second, with a median above 170, and describes the model as “nearly three times faster” than M3.

Artificial Analysis, which evaluates models across a standard set of ten assessments, gave M3.1 Flash an Intelligence Index score of 43. The review says that score is below several open-weight models listed in the comparison, 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 its first token—below MiniMax’s speed claim. The review also reports that it used 180 million tokens to complete the index, compared with a 99 million-token median.

The newsroom test used 12 real news sources in English and German, producing 24 articles with a production prompt. Two judges, GPT-6.1 Sol and Claude Opus 5.5, scored each article from 1 to 5 for source fidelity, invented facts, clarity, usefulness, structure and polish. The test ran through OpenRouter’s free endpoint with a 32,000-token output limit. The supplied source ends before giving the results table, so it does not establish how M3.1 Flash performed or how it compared with other models.

Price and Accuracy Both Matter

For newsrooms and other organizations using AI to turn source material into finished copy, price per token is only part of the cost. A model that consumes more tokens to complete work may reduce the savings from a low listed rate; the review’s reported 180 million tokens on the index, against a 99 million median, makes that distinction relevant. It does not, on its own, show what a newsroom would spend on a particular workload.

Accuracy and faithful use of sources matter at least as much. The newsroom test was designed to examine whether the model invents details or distorts reporting, but its scores are missing from the available material. Readers therefore cannot use this account to decide whether M3.1 Flash is suitable for publication workflows. The independent index and measured speed provide useful evidence about capability, but they do not substitute for the unreported newsroom findings.

From Space Bunny to M3.1

Before its public identification, the model appeared on OpenRouter under the name Space Bunny Alpha. According to the review, it appeared anonymously on September 23, led OpenRouter’s usage ranking for ten days and processed 56.7 trillion tokens on that platform before being removed on October 5. MiniMax says the model processed more than 200 trillion tokens overall. These are platform and company figures reported in the source, not independently established measures of quality.

MiniMax released M3.1 Flash on October 19 through its API, M Plan subscription, MiniMax Code and OpenRouter. OpenRouter offered access free for the first 72 hours. The model has a stated one-million-token context window, adjustable reasoning effort and cached input priced at $0.015 per million tokens. Its weights had not been released as of the October 20 report; MiniMax said they would arrive October 23.

MiniMax also published benchmark comparisons, but the review notes that the company selected the competitors and that its charts omit some newer versions. The report identifies DeepSeek V4.1 Flash 1017, released two days before M3.1 Flash at the same per-token price, as a direct rival. It says DeepSeek led on DeepSWE and Terminal-Bench, while M3.1 Flash scored higher on Automation-Bench. Those comparisons are company-published results; the review says they have not been independently reproduced.

Scores, Weights and Data Questions

The source material does not include the newsroom test’s results table, despite introducing a comparison of open-weight models and closed references. It gives no scores for M3.1 Flash, no rival scores and no verdict from the two judges. No conclusion about its newsroom performance can be drawn from the method alone.

The weights were still pending, and MiniMax had not published a license or official parameter count. The review mentions an outside report estimating 428 billion total parameters and 23 billion active parameters, but says MiniMax had not confirmed those figures. MiniMax’s benchmark and security claims, including a reported CyberGym result of 82%, are described as internal evaluations without independent reproduction in the source.

There is also an unresolved data question from the anonymous OpenRouter period. The report says the provider could retain prompts during that time but was not allowed to train on them. MiniMax has not said what happened to those retained prompts. The report does not establish whether any particular user’s data was retained or how long it may have been stored.

Watch the Weight Release

The next stated milestone is October 23, 2026, when MiniMax said it would release M3.1 Flash’s weights. Their availability, license terms and any official parameter information will affect whether independent researchers can inspect or run the model outside MiniMax’s hosted services.

For a fuller assessment of newsroom use, readers would need the missing 24-article results and the scores for the comparison models, along with details sufficient to interpret the judges’ ratings. MiniMax’s prompt-data explanation for the anonymous OpenRouter period is also outstanding. Until those details are available, the documented evidence supports a view of the model’s price and selected external measurements—not a firm conclusion about its editorial reliability.

Key Questions

What is MiniMax M3.1 Flash?

M3.1 Flash is a model released by MiniMax on October 19, 2026, and marketed for agent work. The report lists it on MiniMax’s API, M Plan, MiniMax Code and OpenRouter.

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. These are token rates, not a guaranteed price for completing a particular task.

Did the newsroom test show that M3.1 Flash writes accurate articles?

The available source does not provide the test scores or comparative results. It describes a 24-article evaluation in English and German but does not support a conclusion about the model’s newsroom accuracy.

When will the model weights be released?

MiniMax said the weights would arrive October 23, 2026. The October 20 report says they had not yet been released, and it does not give a license or confirmed parameter count.

What happened to prompts sent to Space Bunny Alpha?

The report says the anonymous OpenRouter provider could retain prompts but could not use them for training. MiniMax has not clarified what happened to that data, including whether it was retained or deleted.

Source: ThorstenMeyerAI.com