# MiniMax-Text-01

> Open-weight 456B-total / 45.9B-active hybrid-MoE LLM that scales lightning (linear) attention to a 4M-token inference context.

- Maker: MiniMax
- Line: MiniMax-01 (Text-01)
- Canonical page: https://ai-tldr.dev/models/minimax-text-01/
- Updated: 2026-06-18

## In short

MiniMax-Text-01 is the foundational text large language model in MiniMax's MiniMax-01 series, open-sourced on 15 January 2025 alongside the vision-language model MiniMax-VL-01. It is a Mixture-of-Experts model with 456 billion total parameters, of which 45.9 billion are activated per token across 32 experts (Top-2 routing) over 80 layers. It has a 4M-token context window and costs $0.20 per million input tokens and $1.10 per million output tokens.

## Facts

| Fact | Value | Source |
| --- | --- | --- |
| Context window | 4M (max output Not separately published) | [1] |
| Input price | $0.20 / 1M tokens | [2] |
| Output price | $1.10 / 1M tokens | [2] |
| Licence | MiniMax Model License Agreement (open weights, custom) · open weights | [1] |
| Released | 15 January 2025 | [3] |
| Parameters | 456B total / 45.9B active (MoE, 32 experts, Top-2 routing, 80 layers) | [1] |
| Knowledge cutoff | Not officially stated | [1] |

## Sources

1. [Hugging Face model card (MiniMax-Text-01)](https://huggingface.co/MiniMaxAI/MiniMax-Text-01)
2. [MiniMax-01 open-source announcement](https://www.minimax.io/news/minimax-01-series-2)
3. [MiniMax — official site](https://www.minimax.io/)

## Overview

MiniMax-Text-01 is the foundational text large language model in MiniMax's MiniMax-01 series, open-sourced on 15 January 2025 alongside the vision-language model MiniMax-VL-01. It is a Mixture-of-Experts model with 456 billion total parameters, of which 45.9 billion are activated per token across 32 experts (Top-2 routing) over 80 layers. MiniMax positions it as the first time linear ("lightning") attention has been scaled to a commercial-grade model of this size.

Its defining trait is a hybrid attention stack: within every block of 8 layers, 7 use lightning (linear) attention and 1 uses standard softmax attention. That mix keeps the cost of processing very long inputs close to linear, which is how MiniMax-Text-01 reaches a 1-million-token context during training and extrapolates to up to 4 million tokens at inference — among the longest contexts of any model at release. MiniMax reports 100% accuracy on a 4M-token Needle-In-A-Haystack retrieval test and a 0.910 RULER score at 1M tokens.

On core academic benchmarks MiniMax-Text-01 lands in the same range as GPT-4o and Claude 3.5 Sonnet — for example 88.5 on MMLU, 94.8 on GSM8K, and 89.1 on both IFEval and Arena-Hard. The weights are released openly under MiniMax's custom Model License Agreement on Hugging Face and GitHub (code is MIT), and MiniMax recommends vLLM for production serving. It is text-only; multimodal input lives in the separate MiniMax-VL-01. The later MiniMax-M1 reasoning model is built on this same Text-01 base.

## Benchmarks

- MMLU: 88.5% ([source](https://github.com/MiniMax-AI/MiniMax-01/blob/main/README.md))
- MMLU-Pro: 75.7% ([source](https://github.com/MiniMax-AI/MiniMax-01/blob/main/README.md))
- GPQA Diamond: 54.4% ([source](https://github.com/MiniMax-AI/MiniMax-01/blob/main/README.md))
- GSM8K: 94.8% ([source](https://github.com/MiniMax-AI/MiniMax-01/blob/main/README.md))
- MATH: 77.4% ([source](https://github.com/MiniMax-AI/MiniMax-01/blob/main/README.md))
- IFEval (avg): 89.1% ([source](https://github.com/MiniMax-AI/MiniMax-01/blob/main/README.md))
- Arena-Hard: 89.1% ([source](https://github.com/MiniMax-AI/MiniMax-01/blob/main/README.md))
- C-SimpleQA: 67.4% ([source](https://github.com/MiniMax-AI/MiniMax-01/blob/main/README.md))
- LongBench v2 (overall, with CoT): 56.5% ([source](https://github.com/MiniMax-AI/MiniMax-01/blob/main/README.md))
- RULER (1M tokens): 0.91% ([source](https://github.com/MiniMax-AI/MiniMax-01/blob/main/README.md))

## Pricing

- Input: $0.20 / 1M tokens per 1M tokens
- Output: $1.10 / 1M tokens per 1M tokens

MiniMax's official open-source announcement lists $0.20 per million input tokens and $1.10 per million output tokens; OpenRouter lists the same $0.20 in / $1.10 out for minimax/minimax-01. Weights are also free to download and self-host under MiniMax's Model License Agreement.

Source: https://www.minimax.io/news/minimax-01-series-2

## Strengths

- Extremely long context: 1M-token training window that extrapolates to up to 4M tokens at inference — among the longest available at release
- Lightning (linear) attention keeps long-context cost near-linear instead of quadratic, unlike a pure softmax transformer
- Strong long-context retrieval: 100% on a 4M-token Needle-In-A-Haystack test and 0.910 RULER at 1M tokens (per MiniMax)
- Competitive general benchmarks against closed frontier models of its era (MMLU 88.5, GSM8K 94.8, IFEval 89.1, Arena-Hard 89.1)
- Open weights with a low hosted API price ($0.20 in / $1.10 out per 1M tokens)
- Serves as the open base for the MiniMax-M1 reasoning model

## Best for

- Long-document and whole-codebase analysis that needs hundreds of thousands to millions of tokens of context
- Retrieval over very large inputs (long PDFs, transcripts, logs) where needle-in-a-haystack accuracy matters
- Self-hosted general-purpose chat and instruction-following where open weights are required
- A base model for fine-tuning or for building reasoning systems (as MiniMax did with M1)
- Cost-sensitive long-context API workloads via MiniMax or OpenRouter

## How to access

- [MiniMax](https://www.minimax.io/platform_overview) — model id `MiniMax-Text-01`
- [OpenRouter](https://openrouter.ai/minimax/minimax-01) — model id `minimax/minimax-01`

## MiniMax-01 (Text-01) versions

- MiniMax-Text-01 (15 January 2025) — this page

## FAQ

### How large is the MiniMax-Text-01 context window?

MiniMax-Text-01 is trained at a 1-million-token context and can extrapolate to up to 4 million tokens at inference — among the longest contexts available when it launched. Its lightning (linear) attention is what makes processing such long inputs affordable; MiniMax reports 100% accuracy on a 4M-token Needle-In-A-Haystack retrieval test and a 0.910 RULER score at 1M tokens.

### What architecture does MiniMax-Text-01 use?

It is a Mixture-of-Experts model with 456 billion total parameters and 45.9 billion activated per token across 32 experts (Top-2 routing) over 80 layers. Its attention stack is hybrid: within every 8 layers, 7 use lightning (linear) attention and 1 uses standard softmax attention, which keeps long-context cost close to linear.

### Is MiniMax-Text-01 open source, and what license applies?

The weights are openly downloadable on Hugging Face and GitHub under MiniMax's custom Model License Agreement (the code is MIT-licensed). It allows self-hosting and commercial use but adds conditions — for example, naming and attribution requirements and a restriction on using outputs to improve other large language models — so it is open-weight rather than a standard OSI license like Apache 2.0.

### Does MiniMax-Text-01 support images or audio?

No. MiniMax-Text-01 is text-only. Image understanding is handled by the separate MiniMax-VL-01 vision-language model, which adds a vision encoder on top of the same MiniMax-01 base.

## Links

- [MiniMax — official site](https://www.minimax.io/)
- [MiniMax-01 open-source announcement](https://www.minimax.io/news/minimax-01-series-2)
- [Hugging Face model card (MiniMax-Text-01)](https://huggingface.co/MiniMaxAI/MiniMax-Text-01)
- [GitHub repository (MiniMax-01)](https://github.com/MiniMax-AI/MiniMax-01)
- [Technical report — MiniMax-01: Scaling Foundation Models with Lightning Attention (arXiv 2501.08313)](https://arxiv.org/abs/2501.08313)
- [MiniMax API platform](https://www.minimax.io/platform_overview)
- [Benchmark results (GitHub README)](https://github.com/MiniMax-AI/MiniMax-01/blob/main/README.md)
- [OpenRouter (MiniMax-01) — try it](https://openrouter.ai/minimax/minimax-01)

Source: AI/TLDR LLM registry, https://ai-tldr.dev/models/minimax-text-01/