Skip to content

MiniMax-Text-01

A large language model with a 4M token context

Visit official site

What MiniMax-Text-01 is

The model is described as a large language model developed by MiniMaxAI that mixes lightning attention, softmax attention and mixture of experts in one architecture.

What you can do with it

  • Generate long-form text content
  • Work with a 4 million token context
  • Study language understanding and generation

Model notes

  • Total parameters are listed as 456 billion
  • Training context is extended to 1 million tokens
  • Methods named include LASP+, varlen ring attention and ETP
  • An int8 quantisation setting is recommended

Pros & cons

✓ What we like

  • Parallel strategies such as LASP+ and ETP are named
  • Quantisation with int8 is recommended in the steps

! What to watch out for

  • Loading requires device mapping across GPUs
  • Benchmark results are described for academic tests

FAQ

How large is the model?

The material states 456 billion total parameters and 45.9 billion activated per token.

How is it loaded?

The steps name loading config and tokenizer, setting quantisation, mapping devices and preprocessing text.

Last reviewed: 2026-11-15

More AI coding tools tools

View all →

How we review