How much does M3.1 Flash cost to use?

Input costs a flat 15 cents per million tokens, output costs 60 cents per million tokens, and cached input costs 1.5 cents.

What benchmarks did MiniMax report for M3.1 Flash?

It scored 61.7 on OSWorld V2, 76.2 on BankerToolBench, 82% on CyberGym and 73.7% on DeepSWE v1.1. MiniMax said none of these results has been independently checked.

Where can developers use M3.1 Flash now?

It is available through the MiniMax API, the company's M Plan subscription and its MiniMax Code desktop app. The weights are set to be released later.

Large Language ModelsAI Business & IndustrySiliconANGLE AI3m agoOct 20, 2026, 10:01 JST

MiniMax launches M3.1 Flash, cuts AI agent costs by 70%

Send on LINE

3 Key Points

What happened
MiniMax Group Inc. released M3.1 Flash, an update to its June M3 model. The company says agentic tasks cost 70% less and inference runs close to three times as fast, with 428 billion parameters.
Why it matters
AI agents often need dozens of model calls to finish one job, so a 70% per-task cost cut could make always-on agent deployments cheaper to run, though the figures come from MiniMax's own testing.

WHO IT HITSEnterprise teams building AI agents that make many model calls per job — for code auditing, penetration testing or banking workflows — get a lower per-call price and faster responses. Whether the claimed savings hold outside MiniMax's own tests remains unverified.

Not sure about something? Ask the AI

Questions and answers are published on this page.

Summaries like this, in your inbox every morning.

Get news like this every morning, free

Context & Analysis

MiniMax is using M3.1 Flash to compete on price and speed rather than raw size. The model lists 428 billion parameters, barely changed from M3, and the company credits the extra speed to more aggressive sparsity at inference time — meaning fewer parts of the model are engaged per request. Reasoning stays on for every request, with developers picking one of five effort levels from Low to Max, which lets them trade cost against capability.

The launch lands directly on DeepSeek's V4.1-Flash, which charges the same 15 cents per million input tokens off-peak. On MiniMax's own reported figures the two are close on software engineering: M3.1 Flash managed 73.7% on DeepSWE v1.1 against DeepSeek's stated 77%, and neither result has been independently checked. MiniMax's Hong Kong shares fell after the original V4.1-Flash launched in September, according to Bloomberg, so the pricing pressure behind this release is already visible in its stock.

The launch also closes a small mystery. Space Bunny, an anonymous model that appeared on OpenRouter's routing service in late September, was confirmed today as an early build of M3.1 Flash; more than 200 trillion tokens went through it while its maker stayed unnamed. Beyond that confirmation, MiniMax has not detailed the reasoning behind the 200 tokens-per-second peak output figure or how the Pro model will differ from Flash.

FAQ

How much does M3.1 Flash cost to use?: Input costs a flat 15 cents per million tokens, output costs 60 cents per million tokens, and cached input costs 1.5 cents.
What benchmarks did MiniMax report for M3.1 Flash?: It scored 61.7 on OSWorld V2, 76.2 on BankerToolBench, 82% on CyberGym and 73.7% on DeepSWE v1.1. MiniMax said none of these results has been independently checked.
Where can developers use M3.1 Flash now?: It is available through the MiniMax API, the company's M Plan subscription and its MiniMax Code desktop app. The weights are set to be released later.

SiliconANGLE AIRead Original Article

Send on LINE

AI news that matters for your work, delivered every morning.

Pick your industry and the AI tools you use, and get news related to your work every day.

Get it free

Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →

Ask AI

Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.