What better way to start the new year than with another free AI coding model in Kilo? We are excited to announce that MiniMax-M2.1 is now available to use for FREE in Kilo Code.
Popular OSS lab MiniMax is phasing out the free version of their previous model, M2, which has been trending at #5 in Code Mode in Kilo for several weeks now. As a special new years gift to our users, they're offering M2.1, a faster and more powerful upgrade, free for anybody to use in Kilo for a limited time.
Check it out and let us know what you think!
The Future is Open
This release marks a significant moment for open-weight models. MiniMax M2.1 generated almost 200k downloads on Hugging Face and has generated serious buzz in the developer community.
With its massive token context window and efficient Mixture-of-Experts (MoE) architecture (230B total parameters with only 10B active), it has been hailed by some early adopters as a potential "RAG Killer."
Source: Hugging Face
Developers are reporting that M2.1 punches well above its weight class, delivering coding performance that rivals top-tier proprietary models like Claude 3.5 Sonnet and Gemini 3 Pro, particularly in web and mobile app development tasks. Code Arena has been comparing M2.1 against top SOTA models. Its high scores on the VIBE and SWE-bench benchmarks back up these claims, proving that open-source models are no longer just catching up---they are competing at the frontier.
This shift aligns perfectly with what we've been seeing at Kilo. Chinese open-weight models are getting exponentially better. As we noted in our recent blog post on the LLM Convergence:
"OSS models have started to close the gap with SOTA models. How might we track what comes next?"
MiniMax M2.1 is exactly "what comes next."
It represents a new tier of accessibility and power for developers who want state-of-the-art reasoning without the proprietary lock-in. And the focus on multi-lingual capabilities shows an embrace of developer communities around the world.
In our own testing detailed in our MiniMax M2.1 deep dive, we found it to be a powerhouse for agentic workflows. As we highlighted, this model is made for real-world development, excelling at complex debugging and multi-step reasoning.
Starting today, you can experience this performance firsthand in Kilo without denting your AI coding budget at all. Whether you are refactoring legacy code, running a super-fast code review, or building something new in Kilo's App Builder, MiniMax-M2.1 is ready to work.
We're committed to hosting a range of free models on Kilo and grateful to MiniMax for helping that happen.
Try M2.1 in Kilo today!
Top comments (0)