Skip to main content
Rase AI Intelligence Newsletter

Models & assistants Model or AI service

MiniMax-M1 publishes reasoning models for long text

MiniMax released M1 with two downloadable checkpoints. The company reports support for a one-million-token context and documents two different thinking budgets. [2][3]

Model access
Downloadable M1-40K and M1-80K checkpoints. MiniMax also described hosted access at launch. [2][3]
License
Apache License 2.0 for the documented 80K checkpoint. [1][3]

What the model does

The June 2025 announcement introduced the M1-40K and M1-80K checkpoints. MiniMax described the names as the versions’ 40K and 80K thinking budgets. [2][3]

Why it matters

Long context can let a team handle several related documents in one request. Distinct reasoning budgets make it possible to compare latency and answer quality while keeping much of the model family the same. [2][3]

Who can use it

Researchers and teams with suitable serving infrastructure can download the checkpoints and test long-document workloads. MiniMax offered hosted access at launch as well. [2][3]

Key dates

  1. MiniMax-M1 introduced [2]

Ways to use it

Give the model a set of documents with relevant details placed far apart. Check that its answer cites the right passage, and test the same workload with an input length you actually need. [2][3]

Model limitations

The published context limit and the thinking budgets describe different constraints. Reaching the largest documented context in a deployment depends on its serving software and hardware. [2][3]

Sources and verification

Catalog evidence checked How we verify records

  1. Primary source MiniMax’s Apache 2.0 license for its M1-80K checkpoint
  2. Primary source MiniMax’s original M1 release and variant details
  3. Primary source MiniMaxAI/MiniMax-M1-80k
  4. Other source MiniMax-M1: Scaling Test-Time Compute Efficiently with Lightning Attention

Related profiles

Models & assistants

GPT-6.2 Astra and Sol

OpenAI

Astra targets demanding projects. Sol targets everyday tasks. Both support text and image input in the API.

High confidenceEvidence checked

Models & assistants

MAI Code 2 and Code 2 Flash

Microsoft AI

Code 2 targets complex software work. Flash targets frequent edits, explanations and routine fixes.

High confidenceEvidence checked

Models & assistants

MAI Realtime 1

Microsoft AI

Users can try a MAI Labs demo. Developers can register for a limited Foundry preview.

High confidenceEvidence checked