models/minimax/minimax-m2-her
M
minimax·active

MiniMax: MiniMax M2-her

minimax's efficiency model. Context window: 66K tokens.

Overall score
2.83
/5.00 · ranked #145
Input
$0.300
per 1M tokens
Output
$1.20
per 1M tokens
Context
66K
tokens
Blended
$0.975
3:1 out:in ratio

Price drops, new benchmarks, model updates. Stay current on MiniMax: MiniMax M2-her.

One email per change. Unsubscribe anytime.

modelpicker.aipowered by live benchmark data

Scores by test

Methodology →
Structured Output
4.0
Strategic Analysis
1.0
Constrained Rewriting
1.0
Creative Problem Solving
2.0
Tool Calling
Faithfulness
5.0
Classification
4.0
Long Context
4.0
Safety Calibration
1.0
Persona Consistency
4.0
Agentic Planning
3.0
Multilingual
3.0
Tabular Data
2.0

What you need to know

MiniMax M2-her is a specialized model optimized for high faithfulness and structured data handling. It achieves a perfect 5/5 score in faithfulness and strong 4/5 ratings in classification, structured output, and persona consistency. These metrics indicate the model is reliable for extracting information accurately and adhering to specific output formats without introducing hallucinations.

The model struggles significantly with complex reasoning and precision editing. It scores a 1/5 in strategic analysis and constrained rewriting, meaning it cannot handle high-level planning or strict linguistic constraints. Additionally, a 1/5 in safety calibration suggests it may lack the robust guardrails found in more general-purpose frontier models.

At a blended cost of $0.975/MTok, the pricing is competitive, but the overall utility is limited. With an overall rank of 128 out of 130 models and an average internal score of 2.83/5.0, the model provides low general intelligence relative to its cost. Its 66K context window is functional, supporting a 4/5 score in long context tasks, but it lacks the versatility required for general-purpose assistant roles.

Use this model if you need a cost-effective tool for classification, structured data extraction, or maintaining a consistent persona. Skip this model if your use case requires strategic reasoning, complex rewriting, or strict safety compliance.

Strengths — Top 3

Faithfulness5.0/5.0
Structured Output4.0/5.0
Classification4.0/5.0

Relative weaknesses — Bottom 3

Strategic Analysis1.0/5.0
Constrained Rewriting1.0/5.0
Safety Calibration1.0/5.0

Similar models

OGPT-4o$8.133.46MCodestral 2508$0.7503.46MLlama 4 Scout$0.2503.31MLlama 3.3 70B Instruct$0.2653.46