Where MiniMax M3.1 Flash Stands: Leaderboard Rankings And Our Newsroom Test

🔍 Read the full analysis: Where MiniMax M3.1 Flash Stands: Leaderboard Rankings And Our Newsroom Test on ThorstenMeyerAI.com

Buying for a business?Offer from Amazon

Get business pricing on office and shipping supplies

• Business-only prices and quantity discounts
• Tax-exempt purchasing
• Multiple users, one account, clear invoices

Create a free Business account →Free to register

As an affiliate, we earn on qualifying purchases.Hide

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, as it positions the system for agent tasks. An independent index gave it an Intelligence Index score of 43, below several open-weight competitors, while a newsroom evaluation covered 24 articles but its results are not present in the supplied report excerpt. The weights are scheduled for release on October 23, leaving the model’s open-weight status and license unresolved for now.

MiniMax says M3.1 Flash is available through its API, M Plan subscription, MiniMax Code and OpenRouter. OpenRouter offered access free for the first 72 hours. The listed price is $0.15 per million input tokens, $0.60 per million output tokens and $0.015 per million cached input tokens. Those rates are half the input and output prices listed for its predecessor, M3. MiniMax also claims the new model costs about 70% less per agent task, a company comparison that was not independently verified in the report.

The model has a one-million-token context window, adjustable reasoning effort and an OpenRouter output limit of up to 524,288 tokens. MiniMax describes its design as using “more aggressive sparsity” and claims everyday speed above 200 tokens per second, with a median above 170. Artificial Analysis measured 120 tokens per second and 1.08 seconds to first token in its evaluation. The difference reflects distinct measurements and does not, by itself, establish why the results diverge.

The report says M3.1 Flash was previously available anonymously on OpenRouter as Space Bunny Alpha from September 23 to October 5. It led OpenRouter’s usage ranking for 10 days and processed 56.7 trillion tokens on that platform, according to the report; MiniMax says its overall processing exceeded 200 trillion tokens. During the anonymous period, the provider could retain prompts but could not use them for training. MiniMax has not clarified what happened to that retained data.

Price Versus Measured Performance

M3.1 Flash’s low listed token price may appeal to developers running agent workflows, where repeated model calls can add to operating costs. But price per token is only one part of total cost: a model that uses more tokens to complete a task may narrow the savings. Artificial Analysis reported that M3.1 Flash used 180 million tokens across its index evaluation, compared with a 99 million median. That is a measurement from the index’s test, not a direct estimate of what every customer’s workload will require.

The gap between MiniMax’s speed claim and Artificial Analysis’ measured speed also matters to teams comparing latency. The index placed M3.1 Flash below several open-weight models on its Intelligence Index, including DeepSeek V4.1 Flash 1017, which scored 45 against M3.1 Flash’s 43. The figures do not settle which model will work best for a given application, but they give buyers a reason to compare task quality, token use and response time alongside posted prices.

For publishers and other users relying on models to turn source material into finished copy, benchmark scores are not a substitute for checking factual fidelity. The report describes an evaluation based on 24 English- and German-language articles and two judges, but the supplied source ends before presenting the results. No conclusion about M3.1 Flash’s newsroom quality can be drawn from the available details.

From Space Bunny to M3.1

The anonymous OpenRouter model known as Space Bunny Alpha appeared on September 23 and was removed on October 5, according to the source report. MiniMax identified it on October 19 as an early version of M3.1 Flash. The report says the model led OpenRouter usage for 10 days during that period. That history attracted attention before the official release, but usage volume does not establish accuracy or capability.

MiniMax’s published benchmark charts compare M3.1 Flash with several competitors, including DeepSeek V4.1 Flash 1017, Kimi K3, GLM-5.3, GPT-5.6 Sol and Claude Opus 5. In the chart transcription cited by the report, M3.1 Flash scored 62.2 on Automation-Bench, ahead of DeepSeek’s 60.5, but trailed DeepSeek on DeepSWE and Terminal-Bench. The report notes that MiniMax selected the comparison models and that some newer rivals were not included. These results are company-published comparisons, not independent reproductions.

Artificial Analysis classified M3.1 Flash as proprietary because its weights had not yet been released. MiniMax says the weights will arrive on October 23, but the source provides no confirmed license or official parameter count. A parameter estimate reported by another outlet is explicitly unconfirmed by MiniMax and is not treated here as established fact.

““more aggressive sparsity””

— MiniMax

Weights, Data and Test Scores

The report excerpt does not include the newsroom benchmark’s score table, despite describing its 24-article setup. It says each source was written in English and German and judged by GPT-6.1 Sol and Claude Opus 5.5 on source fidelity, invented facts, clarity, usefulness, structure and polish. The scores and comparative results are missing, so it is not possible to say how M3.1 Flash performed or whether it outscored other models in that test.

Other open questions include whether MiniMax will release the weights on October 23, what license will apply and what the model’s official parameter count is. The company has also not said what happened to prompts retained during the anonymous OpenRouter period. MiniMax’s benchmark and security claims, including a reported 82% score on CyberGym, have not been independently reproduced in the source account.

Scheduled Weight Release

The next stated milestone is October 23, 2026, when MiniMax says it plans to release M3.1 Flash’s weights. The release, if it occurs as scheduled, should clarify whether developers can access and run the model outside MiniMax’s hosted services. Its practical openness will also depend on the license and the files made available.

Further assessment will require the missing newsroom scores and independent testing that makes its methods and comparisons clear. Readers evaluating the model can meanwhile distinguish the company’s claims from Artificial Analysis’ measured results and check prices against their own workloads. The report does not provide enough information to establish the newsroom test’s winner or to confirm MiniMax’s claims about per-task savings.

Key Questions

What is MiniMax M3.1 Flash?

M3.1 Flash is a MiniMax model released on October 19, 2026, and marketed for agent work. It is available through several MiniMax services and OpenRouter, according to the source report.

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. MiniMax’s claim that it costs about 70% less per agent task is a company comparison, not an independently verified result in the report.

How did it rank on the independent index?

Artificial Analysis gave M3.1 Flash an Intelligence Index score of 43. The report lists DeepSeek V4.1 Flash 1017 and GLM-5.3 at 45, among other models scoring above M3.1 Flash.

Did M3.1 Flash do well in the newsroom test?

The source describes a test of 24 English- and German-language articles, judged by two models, but the supplied material does not include the scores. Its performance in that test cannot be determined from the report excerpt.

Are the model weights available?

Not at the time covered by the report. MiniMax said the weights would arrive on October 23, 2026; the source does not confirm their release or specify a license.

Source: ThorstenMeyerAI.com