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 is a speed-first model for tool-using agents, software engineering and desktop control — 200+ TPS with far lower token spend.
All Tools
Discover our comprehensive AI-powered animation toolkit
MiniMax H3
MiniMax H3 AI Video Generator
Seedance 2.5
The Future of AI Video Is Here.
Seedance 2.0
The Future of AI Video Is Here.
FLUX 3 Video Generator
Kling 3.0
Next-Gen AI Video Generator
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.
MiniMax M3.1 Flash — Speed Without Sacrificing Depth
Successor to the stealth Space Bunny deployment, MiniMax M3.1 Flash has been stress-tested on enterprise automation, software engineering, desktop operation and security analysis.
- Multi-Step Agent AutomationSits near the top of AutomationBench-AA and posts 76.2 on BankerToolBench, carrying SaaS processes end to end across sales, marketing, ops, support, finance and HR.
- Software Engineering DepthScores 73.7% on DeepSWE v1.1 by reading unfamiliar code, shipping features, patching defects and verifying its own output with tests over long step chains.
- Desktop Control and Security TestingLogs 61.7 on OSWorld V2 and 29.6 on Agents' Last Exam for live-environment operation, and 82% on CyberGym for finding and confirming vulnerabilities.
Getting MiniMax M3.1 Flash into Production
From your very first call to production-grade agent workloads in three short moves.
What's Under the Hood of MiniMax M3.1 Flash
Quick serving, predictable per-token rates and verified results in automation, software work, desktop control and security engineering.
Deeper Sparsity, Quicker Replies
Extra sparsity across the serving stack pushes routine output past 200 TPS while the P50 stays above 170 TPS — roughly a threefold jump in online inference speed.
Simple, Predictable Token Rates
At $0.15 per million input, $0.60 per million output and $0.015 for cached input, total agent spend lands about 70% under the prior generation.
SaaS Workflows, Start to Finish
A top-tier AutomationBench-AA position plus 76.2 on BankerToolBench reflect multi-step planning, compliance checks and linked tool calls inside live SaaS environments.
Real Software Engineering
Scoring 73.7% on DeepSWE v1.1 places it with the leaders on demanding builds that require grasping a codebase, adding features, closing bugs and validating with tests.
Hands-On Computer Control
With 61.7 on OSWorld V2 and 29.6 on Agents' Last Exam — top among Chinese models on each — it operates live environments, drives tools and completes tasks.
Security and Code Auditing
An 82% CyberGym result, confirmed by in-house expert review, spans code audits, patch work, pen testing, malware triage and reverse engineering at modest cost.
MiniMax M3.1 Flash: Common Questions
Straight answers on cost, speed, benchmark scores and how to get access to this model.
What kind of model is MiniMax M3.1 Flash?
It's MiniMax's newest Flash release, grown out of the stealth Space Bunny deployment that handled more than 200 trillion tokens, with clear gains in agent work, coding, desktop control and security testing.
What do a million tokens cost?
Pricing is flat: $0.15 per million input tokens, $0.60 per million output and $0.015 per million cached input — about 70% cheaper overall than the model it replaces.
Where does the near-3x speedup come from?
It comes from heavier sparsity in the serving stack, which lifts routine throughput to 200+ TPS while the median (P50) holds above 170 TPS.
Which benchmark scores support these numbers?
A leading AutomationBench-AA position, 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 start using it?
Via the MiniMax API on platform.minimax.io, an M Plan subscription, or MiniMax Code — the desktop build is available from agent.minimax.io/download, with open weights arriving later.
How is Flash different from Pro?
Flash is MiniMax's continuing push on the efficiency-versus-capability balance; Pro, which targets the outer edge of raw intelligence, lands later.
Put MiniMax M3.1 Flash to Work Today
Grab API access, an M Plan seat or the MiniMax Code desktop app and get frontier-level agent, coding, desktop and security performance for about 70% less spend at near-3x the speed.