> ## Content Index
> Fetch the complete content index at: https://www.testingcatalog.com/llms.txt
> Use this file to discover other available public pages before exploring further.

# MiniMax unveils MiniMax-01 Series 2 with 456B-parameter AI model
- URL: https://www.testingcatalog.com/minimax-unveils-minimax-01-series-2-with-456b-parameter-ai-model/
- Published: 2025-01-15T00:01:18.000Z
- Updated: 2025-01-15T00:01:18.000Z
- Author: Alexey Shabanov
- Tags: Latest AI News

MiniMax has announced the release of the **MiniMax-01 Series 2**, which includes advancements in their AI model lineup, particularly the **MiniMax-Text-01**. This update highlights their commitment to pushing boundaries in AI technology, focusing on large-scale language models and practical applications.

The **MiniMax-Text-01** is a state-of-the-art Mixture of Experts (MoE) language model with 456 billion parameters, 45.9 billion of which are activated per token. It employs a hybrid attention mechanism combining Lightning Attention and Softmax Attention to optimize performance. The model supports extensive context lengths, with a training capacity of up to 1 million tokens and inference capabilities reaching 4 million tokens. This makes it suitable for tasks requiring deep contextual understanding and long-form text processing. Additionally, its use of Rotary Position Embedding (RoPE) enhances positional encoding for half the attention head dimensions, ensuring efficient processing of complex data structures.

> MiniMax-01 is Now Open-Source: Scaling Lightning Attention for the AI Agent Era  
>  
> We are thrilled to introduce our latest open-source models: the foundational language model MiniMax-Text-01 and the visual multi-modal model MiniMax-VL-01.  
>  
> 💪Innovative Lightning Attention… [pic.twitter.com/LbJhhmxD4P](https://t.co/LbJhhmxD4P?ref=testingcatalog.com)
> 
> — MiniMax (official) (@MiniMax\_\_AI) [January 14, 2025](https://twitter.com/MiniMax%5F%5FAI/status/1879226391352549451?ref%5Fsrc=twsrc%5Etfw&ref=testingcatalog.com)

Key specifications include:

1. **80 layers** with alternating attention mechanisms.
2. **32 experts** in the MoE framework, utilizing top-2 routing.
3. A hidden size of 6144 and a vocabulary size of 200,064 tokens.

The MiniMax-01 Series 2 models demonstrate competitive performance against other leading AI systems like Qwen and DS3, excelling in benchmarks for long-context comprehension. However, some challenges remain in instruction adherence and coding tasks, areas where further refinement is anticipated.

The company behind this innovation, MiniMax, is a Chinese AI startup that has gained significant traction with its consumer-facing app "Talkie," a virtual companion platform that has surpassed competitors like Character.AI in popularity. MiniMax's focus on both enterprise-level AI solutions and consumer applications underscores its versatility in addressing diverse market needs.

💡

Try MiniMax-01 on [Hailuo AI](https://www.hailuo.ai/?ref=testingcatalog.com)

This release positions MiniMax as a formidable player in the AI landscape, particularly as it continues to innovate in large-scale models and practical AI deployments. Industry observers will be watching closely to see how these advancements impact both technical applications and user experiences globally.

[Source](https://www.minimaxi.com/en/news/minimax-01-series-2?ref=testingcatalog.com)