MiniMax officially releases M3.1 Flash, disclosing benchmark results,API pricing, and a plan to October 23 release weights
BIBIBI
AT A GLANCE
MiniMax releases M3.1 Flash, disclosing benchmark results,API pricing, and plans to October 23 release weights on
Article
MiniMax officially releases M3.1 Flash: per million Token lowest 0.15 USD，23 Open-sourced on
Beating Insights AI breaking news: MiniMax officially released M3.1 Flash, confirming that the previously viral anonymous model Space Bunny was its early version. Space Bunny launched on OpenRouter and OpenCode in late September and accumulated more than 200000000000000 Token, once topping the daily call rankings on both platforms. MiniMax subsequently released September 27 M3.1 Flash Preview, and has now officially disclosed its full benchmark results and API pricing.
In tests published by MiniMax,M3.1 Flash achieved 62.2%on the office automation benchmark Automation-Bench, exceeding DeepSeek V4.1 Flash's 60.5%and Opus 5's 50.3%. It scored 73.7%in the DeepSWE programming test, close to Opus 5's 74.0%; it reached 91.0%on the visual reasoning benchmark BabyVision. However, in the computer-operation test OSWorld, its 61.7%still lagged behind Opus 5's 70.6%.
Pricing is also a key focus.API charges Token per million input 0.15 USD, 0.60 USD for output, and only 0.015 USD for cache reads. MiniMax claims that, compared with M3, the new model's overall cost is reduced to about one-third, while its speed is increased to 3 times. The model supports a million Token context, and its weights are planned to be released on October 23.
Original link https://m.theblockbeats.info/flash/372587
Key points
01
MiniMax officially releases M3.1 Flash and confirms that the anonymous model Space Bunny was its early version.
02
The original article states that Space Bunny launched on OpenRouter and OpenCode in late September, accumulated more than 200000000000000 Token, and once topped the daily call rankings on both platforms.
03
MiniMax released September 27 M3.1 Flash Preview。
04
Tests published by MiniMax show that M3.1 Flash scored 62.2%on Automation-Bench, 73.7%on DeepSWE, 91.0%on BabyVision, and 61.7%。
05
API per million input Token charges 0.15 USD, output Token charges 0.60 USD, and cache reads cost 0.015 USD。
06
MiniMax claims that, compared with M3, the new model's overall cost is reduced to about one-third and its speed is increased by 3 times.
07
The model supports a million Token context, and its weights are planned to be released on October 23.
AI-assisted interpretation
The following is analysis, separate from reported facts. Verify important claims independently.
M3.1 Flash is a model released by MiniMax. The report lists its results in office automation, programming, visual reasoning, and computer-operation tests, as well as its charges for input, output, and cache reads.API Some results are higher than those of the comparison models listed in the article, but its OSWorld result is lower than Opus 5。
Why it matters to readers
Lower API pricing, a million Token context, and the planned weight release may affect how users assess price, processing length, and deployment options when choosing a model.
Beginners should focus on actual usage costs, task performance, and whether the weights are released as planned, rather than judging model capabilities solely on individual benchmark scores.
Risks and unknowns
The benchmark results come from tests published by MiniMax; the original article does not provide independent replication results.
The claims that cost is reduced to about one-third and speed is increased by 3 times are MiniMax's claims; the original article does not specify the testing conditions.
The weight release is a future plan; whether it will be implemented on October 23 cannot be confirmed from the original article.
The headline says “open source,” but the body only says “release weights”; the specific license terms and scope of release are not stated.
The original article does not specify the conditions and limitations for the million Token context.
Related Developments
Loading event timeline…
Related concepts
Token
This term is not in the glossary yet. Browse related concepts in the glossary.
