MiniMax M3.1 Flash Workbench
A single flat-rate home for multi-step agents, software builds, desktop control and security audits
None

None

None

Long Story Video Skill

Turn your idea into 10s–10min video—AI writes scripts, prompts, and generates footage automatically.

Ads Video Skill

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

AI Video Prompt Generator

Feedback

freeTrialImage.bannerPity

freeTrialImage.upgradeUnlock

  • ✓freeTrialImage.benefitHd
  • ✓freeTrialImage.benefitWatermark
  • ✓freeTrialImage.benefitUnlimited

minimax m3.1 flash

A fast, low-cost model for multi-step agents, code work and real computer tasks — MiniMax M3.1 Flash streams 200+ tokens per second.

All Tools

Discover our comprehensive AI-powered animation toolkit

MiniMax M3.1 Flash: Built for Long-Horizon Work

Born from the anonymous Space Bunny release, this model was stress-tested on enterprise automation, software engineering, desktop control and security engineering before launch.

  • Multi-Step Agent Automation
    It lands near the top of AutomationBench-AA and posts 76.2 on BankerToolBench, carrying SaaS workflows end to end across sales, marketing, operations, service, finance and HR.
  • Software Engineering at Depth
    A 73.7% score on DeepSWE v1.1 reflects reading unfamiliar code, shipping features, patching defects and checking its own work against tests over long step sequences.
  • Desktop Control and Security Work
    Real-environment operation earns 61.7 on OSWorld V2 and 29.6 on Agents' Last Exam, while CyberGym vulnerability discovery and validation reaches 82%.

Three Steps to Start with MiniMax M3.1 Flash

From your first API call to production-grade agent runs in three short moves with MiniMax M3.1 Flash.

Capabilities Built into MiniMax M3.1 Flash

Quick inference, predictable low token bills and measured strength in automation, software engineering, desktop operation and security review.

Sparser Serving, Speedier Replies

Heavier sparsity in the serving stack pushes everyday throughput past 200 TPS while P50 holds above 170 TPS — close to three times quicker than the prior release.

One Flat Rate for Tokens

At $0.15 per 1M input, $0.60 per 1M output and $0.015 per 1M cached input, total agentic spend falls roughly 70% versus the previous model.

SaaS Workflows, End to End

A leading AutomationBench-AA position and 76.2 on BankerToolBench come from multi-step planning, financial compliance and chained tool calls inside live SaaS systems.

Complex Code Work

73.7% on DeepSWE v1.1 places it alongside top models on software jobs that call for independent code reading, feature builds, bug fixes and test validation.

Hands-On Computer Control

61.7 on OSWorld V2 and 29.6 on Agents' Last Exam — best among Chinese models on both — show it operating real environments, calling tools and finishing tasks.

Security and Vulnerability Work

An 82% CyberGym result, backed by internal expert review, spans code auditing, patch work, penetration testing, malware analysis and reverse engineering at low cost.

FAQ

MiniMax M3.1 Flash: Your Questions Answered

Straight answers on cost, speed, benchmark scores and how to get access to MiniMax M3.1 Flash.

1

What kind of model is MiniMax M3.1 Flash?

It is MiniMax's next-generation Flash release, grown out of the anonymously operated Space Bunny system that handled more than 200 trillion tokens; it pushes ahead on agentic tasks, coding, computer use and cybersecurity.

2

What does a million tokens cost?

Rates are flat: $0.15 for 1M input tokens, $0.60 for 1M output tokens and $0.015 for 1M cached input tokens, leaving overall agentic usage roughly 70% cheaper than the earlier model.

3

Where does the near-3x speed gain come from?

Greater sparsity across the serving stack: day-to-day output rises past 200 TPS, and median (P50) speed stays above 170 TPS.

4

Which benchmark scores support these numbers?

AutomationBench-AA top-tier placement, 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.

5

How do I start using it?

Through the MiniMax API at platform.minimax.io, an M Plan subscription and MiniMax Code, whose desktop build is downloadable from agent.minimax.io/download; open weights arrive later.

6

How is Flash different from Pro?

Flash continues MiniMax's exploration of the efficiency-capability trade-off, while Pro — aimed at pushing the frontier of intelligence — lands later.

Put MiniMax M3.1 Flash to Work Today

Call the MiniMax API, take an M Plan subscription or install MiniMax Code and get frontier agent, coding, desktop and security results for about 70% less spend at near-3x the speed.