MiniMax · Released October 19, 2026

MiniMax M3.1
Flash

Built for agents that get work done. Explore faster execution, stronger coding, and a lower cost per task.

200+TPSreported output speed
73.7%DeepSWE v1.1
70%lower agent usage cost*

*MiniMax-reported reduction versus its predecessor for agentic tasks. Benchmark scores and speed figures are provider-reported.

Capabilities

From understanding to execution

Choose a model around the work your agent needs to finish.

01

Software development

A fit for agents that inspect a repository, implement a change, and iterate through test feedback.

02

Connected workflows

Bring planning and tool calls into workflows that span documents, files, and business systems.

03

Computer use & security

Explore desktop automation and security tasks with a model evaluated on real environment interaction.

MiniMax M3.1 Flash benchmarks

MiniMax M3.1 Flash benchmark results

Explore the official results across software engineering, business automation, computer use, and cybersecurity.

MiniMax-reported evaluations. Compare results within the same benchmark and test the model on your own workload.

Official MiniMax M3.1 Flash benchmark comparison
BenchmarkCategoryMiniMax M3.1 FlashDeepSeek V4.1 Flash-1017DeepSeek V4.1 FlashKimi K3GLM 5.3GPT-5.6 SolOpus 5
Automation-Bench v1.0.6CoWork62.860.554.846.748.845.850.3
BankerToolBench76.2—73.769.174.574.676.2
GDP.pdf AA20.8—12.822.011.227.221.6
Terminal Bench 4.038.142.331.212.637.939.951.8
LOCA-Bench 256K60.3——49.750.360.053.7
Agents’ Last ExamComputer Use29.6——28.3—30.632.2
OSWorld 2.061.7—54.358.3—65.770.6
PostTrain Bench v1.1Coding39.3—30.536.637.536.240.4
DeepSWE v1.173.777.074.267.566.973.074.0
NL2Repo-Bench61.268.265.458.058.056.875.3
SWE Refactor Bench23.0—11.819.510.428.531.0
CyberGymCyber82.2——80.084.583.6—
BabyVision (w/tools)General Capabilities91.092.589.685.7—89.994.1
HLE (w/o tools)43.444.039.243.542.349.555.6
CharXiv RQ82.5—69.084.8—85.283.7

Source:MiniMax official announcement ↗· October 19, 2026. Bold values indicate the highest published score in each row. — means no score was reported, not zero. Evaluation setups vary; see the source for methodology.

Inference efficiency

Faster execution. Lower task cost.

MiniMax reports nearly three times faster online inference and 70% lower agentic usage cost than its predecessor. Flash is built for workflows where repeated model and tool calls add up.

What the chart showsOn Automation-Bench, Flash reaches a 62.8% score at a lower plotted cost per task than the other models shown.

200+TPS output speed
170+TPS median (P50)
MiniMax API · per 1M tokens

$0.15 input · $0.60 output · $0.015 cached input

Provider API pricing; separate from MyClaw billing.

MiniMax reports nearly three times faster online inference and 70% lower agentic usage cost than its predecessor. Flash is built for workflows where repeated model and tool calls add up.
Official MiniMax chart
Real-world execution

Built for work beyond the chat

Evaluate Flash on implementation tasks, cross-application actions, and workflows that depend on sustained tool use. The official release covers both coding and computer-operation benchmarks.

What the chart showsOn Agents’ Last Exam, Flash reaches a 29.6% pass rate at an estimated $111 total cost in MiniMax’s evaluation. Some models score higher at a higher plotted cost.

73.7%DeepSWE v1.1
61.7OSWorld V2
29.6Agents’ Last Exam
View the official coding performance / cost chart ↗
On Agents’ Last Exam, Flash reaches a 29.6% pass rate at an estimated $111 total cost in MiniMax’s evaluation. Some models score higher at a higher plotted cost.
Official MiniMax chart
Model comparison

MiniMax M3.1 Flash vs MiniMax M3

Two different starting points for your next agent workflow.

MiniMax model positioning comparison
Model comparisonM3.1 FlashM3
FocusEfficient agent execution and lower task costLong-context coding and multimodal workflows
Speed200+ TPS output; 170+ TPS median, per MiniMaxThe reference model page does not give a comparable speed figure
Context windowNot specified in the launch articleUp to 1M tokens, per the M3 model page
Evaluation highlightsDeepSWE v1.1, OSWorld V2, CyberGymSWE-Bench Pro and long-horizon agent execution
Best first testRepeated tool calls, automation, and cost-sensitive codingLarge repositories, long documents, and multimodal context

Different benchmarks are not directly comparable. Sources:M3.1 Flash announcement ↗andMyClaw M3 model page ↗.

Use cases

Where MiniMax M3.1 Flash fits

Repository work

Inspect code, implement features, and iterate through test feedback in software development workflows.

Business automation

Move from a business task to a structured plan and tool actions across documents, files, and connected systems.

Computer-use workflows

Navigate interfaces and complete multi-step tasks across applications with an agent that acts in real environments.

Security engineering

Investigate code, review potential vulnerabilities, and develop patches in authorized security workflows.

MiniMax M3.1 Flash FAQs

Your next agent workflow

Find your model.
Get to work with MyClaw.

Bring your tasks, files, and tools into a hosted agent workspace. Explore the models currently available for your account.

Explore models in MyClaw