Still a cheap option for developers working in Rust, Go or mobile code, though newer MiniMax models have since arrived.
Kenning's take
What it does
MiniMax M2.1 is a language model aimed at coding and agent work, with improved support for languages such as Rust, Java, Go, C++, Kotlin and TypeScript, and for native Android and iOS development. MiniMax says it gives shorter answers than M2 and works well inside tools such as Claude Code, Cline and Kilo Code. The weights are on Hugging Face, and it is still offered in the MiniMax API, now listed as a legacy model behind newer releases.
- Stronger multi-language coding (Rust, Java, Go, C++, Kotlin and more)
- Native Android and iOS app development
- Interleaved thinking
- 204,800-token context window
- Open weights on Hugging Face
- Highspeed variant at about 100 tokens per second
Who it suits. Developers who want a low-cost or self-hosted model for coding agents.
Key facts
- Category
- AI models
- Made by
- MiniMax
- Released
- 23 December 2025
- How to get it
- through the MiniMax API (MiniMax-M2.1 and MiniMax-M2.1-highspeed), and as open weights on Hugging Face
- Works on
- API, self-hosted
- Free plan
- open weights free to download
- API price
- $0.30 per million input tokens, $1.20 per million output; highspeed $0.60 and $2.40
FAQ
When was MiniMax M2.1 released?
MiniMax announced M2.1 on 23 December 2025.
Is MiniMax M2.1 open source?
Yes, MiniMax published the M2.1 weights on Hugging Face for local use.
How much does MiniMax M2.1 cost?
In the MiniMax API it costs $0.30 per million input tokens and $1.20 per million output tokens.
Facts from the maker's own pages and the sources listed. Checked 2026-10-11.
Similar tools
ACE-Step 1.5
Open-source music generation model that writes full songs with lyrics and runs on ordinary consumer GPUs.
AlphaGenome Atlas
Google DeepMind database of AI-predicted effects for every possible single-letter change in human DNA.
Amazon Nova 2
Amazon's second family of Nova AI models, covering reasoning, speech conversation and combined text-and-image generation.
Cisco Antares
Cisco's small open-weight models that find where known security vulnerabilities sit in a codebase, running locally.