AIToday

Chinese AI startup MiniMax to open-source 2.7 trillion parameter model

THE DECODER3h ago
Chinese AI startup MiniMax to open-source 2.7 trillion parameter model

Key takeaway

Chinese AI startup MiniMax is developing a 2.7 trillion parameter language model planned for open-source release as early as Q3. The model would surpass any existing Chinese AI model in size and compete directly with other open-source offerings, though emerging government controls on such releases may affect the timeline.

Summaries like this, in your inbox every morning.

Sign up free →

3 Key Points

  • What happened

    MiniMax, a Chinese AI developer, is working on a new large language model called M3 Pro with 2.7 trillion parameters and plans to release it as open source, potentially as early as Q3. The model would be larger than any other Chinese AI model currently on the market.

  • Why it matters

    Open-source models from Chinese developers have gained traction with engineers seeking cost-effective options for high-volume, less critical tasks. MiniMax's release would expand competition in this space alongside rivals like Zhipu, DeepSeek, and Moonshot AI, though the Chinese government appears to be considering tighter controls on future open-source model releases.

  • What to watch

    The release could come as early as Q3, though the internal name M3 Pro may change before launch. MiniMax's current top model, M3, has 428 billion parameters, making the new model substantially larger.

Context & Analysis

MiniMax's plan to release a 2.7 trillion parameter open-source model reflects a shift in the Chinese AI market toward larger, publicly available models. The company competes directly with DeepSeek, Zhipu, and Moonshot AI, all of which have seen developer interest in their open-source offerings, particularly for organizations seeking economical solutions for high-volume, lower-stakes applications. The model's size—substantially larger than MiniMax's current M3—suggests the company is pursuing performance gains in complex reasoning tasks to match or exceed rival offerings.

However, the release timeline and scope may face headwinds. Recent reports indicate the Chinese government is considering stricter controls on future open-source model releases, which could affect when and how MiniMax distributes the new model. The name M3 Pro remains internal and may change before launch, leaving room for adjustments to the final product. For developers outside China, the release could widen access to large open-source alternatives, though regulatory constraints on the developer side may ultimately shape adoption.

FAQ

When is MiniMax planning to release the new model?
A release could come as early as Q3, according to sources cited by The Information, though no final launch date has been confirmed.
How does the new model compare to MiniMax's current offering?
The new M3 Pro model has 2.7 trillion parameters, compared with 428 billion parameters in MiniMax's current top model, M3.
What is the advantage of larger models?
Larger models tend to perform better on tasks that require complex reasoning and multi-step instructions.

Discussion

No comments yet. Be the first to share your thoughts!

Log in to join the discussion

Related Articles

Stay ahead with AI news

Get curated AI news from 200+ sources delivered daily to your inbox. Free to use.

Get Started Free

Free · takes 30 seconds · unsubscribe anytime

1 minute a day. The AI essentials.

200+ sources · Email / LINE / Slack

Get it free →