None
None
Long Story Video Skill
Turn your idea into 10s–10min video—AI writes scripts, prompts, and generates footage automatically.
Ads Video Skill
Generate professional ads and sales videos—AI auto-generates scripts, prompts, and footage.
3D Science Explainer Video Skill
Convert scientific concepts into stunning 3D explain animations
Feedback
freeTrialImage.bannerPity
freeTrialImage.upgradeUnlock
- ✓freeTrialImage.benefitHd
- ✓freeTrialImage.benefitWatermark
- ✓freeTrialImage.benefitUnlimited
minimax m3.1 flash
MiniMax M3.1 Flash runs agents, code and computer tasks at blistering speed — over 200 tokens per second for a fraction of the usual token bill.
All Tools
Discover our comprehensive AI-powered animation toolkit
Muse Video
Seedance2.5
The Future of AI Video Is Here.
Seedance 2.0
The Future of AI Video Is Here.
Veo3.1
Create Stunning Videos with Veo3.1
Grok Video Generator
Create Videos from Text or Images with AI
MiniMax H3 video generator
MiniMax H3 AI Video Generator
Seedance 2.0
The Future of AI Video Is Here.
Gemini Omni 1.2 Flash
The Case for MiniMax M3.1 Flash in Real Workloads
Succeeding the quietly released Space Bunny system, MiniMax M3.1 Flash has been stress-tested on enterprise automation, software engineering, desktop control and security tasks.
- Automation That Runs ItselfPlaces near the top of AutomationBench-AA and notches 76.2 on BankerToolBench, completing full SaaS workflows that span sales, marketing, operations, support, finance and HR.
- Software Engineering, End to EndHits 73.7% on DeepSWE v1.1 — reading codebases, shipping features, squashing defects and checking its own work with tests across long multi-step runs.
- Desktop Control and Security AuditsNotches 61.7 on OSWorld V2 and 29.6 on Agents' Last Exam for live-environment operation, plus 82% on CyberGym for finding and confirming vulnerabilities.
Running MiniMax M3.1 Flash in Three Moves
From your very first API call to production-grade agent jobs, here is the entire path laid out.
What You Get Inside MiniMax M3.1 Flash
Quicker responses, predictable flat token rates, and proven results in automation, coding, desktop control and security engineering.
Deeper Sparsity, Quicker Answers
Heavier sparsity in the serving stack pushes everyday throughput past 200 tokens per second with the median holding above 170 — close to a threefold speedup for online inference.
One Flat Rate for Every Token
$0.15 per million input tokens, $0.60 per million output and $0.015 per million cached trims overall agent spend by roughly 70% versus the model it replaces.
SaaS Workflows, Actually Executed
A leading AutomationBench-AA position and 76.2 on BankerToolBench show multi-step planning, financial compliance and chained tool calls running against live SaaS systems.
Hands-On Software Development
73.7% on DeepSWE v1.1 puts the model alongside the front runners on complex engineering that requires reading code, building features, fixing bugs and validating with tests.
Operating in Real Environments
61.7 on OSWorld V2 and 29.6 on Agents' Last Exam — first among Chinese models on both — show it clicking through real apps, using tools and finishing the task.
Backup for Security Engineering
82% on CyberGym, confirmed by in-house expert review, spans code auditing, patch work, penetration testing, malware triage and reverse engineering at low cost.
MiniMax M3.1 Flash: Your Questions Answered
Straight answers on cost, speed, benchmark results and how to get your hands on MiniMax M3.1 Flash.
So what is this model, exactly?
It is MiniMax's next Flash-generation model, grown out of the anonymously run Space Bunny system that handled more than 200 trillion tokens, with clear gains in agentic work, coding, desktop control and cybersecurity.
What will a million tokens set me back?
Pricing stays flat: $0.15 per million input tokens, $0.60 per million output tokens and $0.015 per million cached input — roughly 70% cheaper for agentic use than the model it replaces.
Why is inference nearly three times faster?
Deeper sparsity across the serving stack: routine throughput reaches past 200 tokens per second, and the median (P50) holds above 170.
Which benchmarks support these numbers?
Top placement on AutomationBench-AA, 76.2 on BankerToolBench, 73.7% on DeepSWE v1.1, 61.7 on OSWorld V2, 29.6 on Agents' Last Exam and 82% on CyberGym.
How do I get access?
Sign up for the MiniMax API at platform.minimax.io, take an M Plan subscription, or install MiniMax Code from agent.minimax.io/download; open weights arrive later.
How does Flash differ from Pro?
Flash continues MiniMax's work on balancing efficiency against capability, while Pro — built to push the intelligence frontier — lands at a later date.
Put MiniMax M3.1 Flash to Work Today
Grab an API key, subscribe to M Plan or install MiniMax Code and get frontier-level agent, coding, desktop and security performance — around 70% cheaper, with inference nearly three times quicker.