MiniMax M2.1

Name: MiniMax M2.1
Author: MiniMax

MiniMax

MiniMax M2.1 is an updated version of the M2 language model with improved reasoning, coding, and instruction following. It supports multimodal inputs and delivers strong performance across general knowledge, mathematics, and programming benchmarks.

Key Specifications

Parameters

Context

1.0M

Release Date

December 22, 2025

Average Score

63.6%

API Documentation Model Weights Results Blog

Timeline

Key dates in the model's history

Announcement

December 22, 2025

Last Update

January 29, 2026

Today

September 10, 2026

Technical Specifications

Parameters

Training Tokens

Knowledge Cutoff

Family

Capabilities

MultimodalZeroEval

Pricing & Availability

Input (per 1M tokens)

$0.30

Output (per 1M tokens)

$1.20

Max Input Tokens

1.0M

Max Output Tokens

100.0K

Supported Features

Function CallingStructured OutputCode ExecutionWeb SearchBatch InferenceFine-tuning

Benchmark Results

Model performance metrics across various tests and benchmarks

Programming

Programming skills tests

SWE-Bench Verified

mini-swe-agent • Self-reported

67.0%

Reasoning

Logical reasoning and analysis

GPQA

Accuracy • Self-reported

81.0%

Other Tests

Specialized benchmarks

VIBE Web

Standard evaluation • Self-reported

91.5%

VIBE Android

Android- • Self-reported

90.0%

VIBE

Web, Simulation, Android, iOS, Backend • Self-reported

89.0%

MMLU-Pro

Accuracy • Self-reported

88.0%

VIBE iOS

iOS- • Self-reported

88.0%

VIBE Simulation

Standard evaluation • Self-reported

87.0%

AA-LCR

• Self-reported

62.0%

AIME 2025

• Self-reported

81.0%

BrowseComp

context management • Self-reported

62.0%

Humanity's Last Exam

without tools • Self-reported

22.0%

IFBench

• Self-reported

70.0%

LiveCodeBench

• Self-reported

78.0%

Multi-SWE-Bench

• Self-reported

49.4%

OctoCodingBench

• Self-reported

26.1%

SciCode

• Self-reported

39.0%

SWE-bench Multilingual

• Self-reported

72.5%

SWE-Perf

• Self-reported

3.1%

SWE-Review

• Self-reported

8.9%

SWT-Bench

• Self-reported

69.3%

Tau2 Telecom

• Self-reported

87.0%

Terminal-Bench

Terminal-bench 2.0 • Self-reported

47.9%

Toolathlon

• Self-reported

43.5%

VIBE Backend

• Self-reported

86.7%

License & Metadata

License

mit

Announcement Date

December 22, 2025

Last Updated

January 29, 2026

Similar Models

All Models

MiniMax M2.7

MiniMax

Released:Mar 2026

Price:$0.30/1M tokens

MiniMax M2

MiniMax

230.0B

Best score:0.8 (GPQA)

Released:Oct 2025

Price:$0.30/1M tokens

Solar Pro 4

Upstage

Best score:0.9 (GPQA)

Released:Aug 2026

Price:$0.30/1M tokens

Qwen3.7 Max

Alibaba

Best score:0.9 (GPQA)

Released:May 2026

Price:$2.50/1M tokens

o1-mini

OpenAI

Best score:0.9 (HumanEval)

Released:Sep 2024

Price:$3.00/1M tokens

o1

OpenAI

Best score:0.9 (MMLU)

Released:Dec 2024

Price:$15.00/1M tokens

o1-preview

OpenAI

Best score:0.9 (MMLU)

Released:Sep 2024

Price:$15.00/1M tokens

KAT-Coder-Pro V1

KwaiKAT

Released:Mar 2026

Recommendations are based on similarity of characteristics: developer organization, multimodality, parameter size, and benchmark performance. Choose a model to compare or go to the full catalog to browse all available AI models.