Command Palette
Search for a command to run...

MiniMax Builds Open Source AI Model With 2.7 Trillion Parameters for Q3 Release

aiai-modelingai-model-releasesai-open-models 4 posts · 4 accounts

Chinese AI developer MiniMax is building a language model with 2.7 trillion parameters, dubbed M3 Pro, with an open source release targeted for the third quarter, according to sources familiar with the project.

The system is the largest open weight model currently available from China, exceeding the parameter size of MiniMax’s existing M3 version by more than six times. Competitors including Kimi and Zhipu have recently expanded their open source capabilities, underscoring a sustained push by Chinese developers to benchmark large foundation models.

From the sources (4 posts)

@kimmonismus

Huge: China’s MiniMax Plans to Launch 2.7-Trillion Parameter Model (MiniMax Pro) tl;dr MiniMax is preparing a 2.7T parameter open-source model, potentially launching as early as Q3. That would make it far larger than any Chinese model cur

@teortaxestex

Minimax needs to flip the narrative. With 2.7T it should be possible. Go on, surprise me. Kimi did it, Zhipu did it. Can you?

@techmeme

Sources: Chinese AI developer MiniMax is working on a 2.7T-parameter model, dubbed M3 Pro and set for release as early as Q3, and plans to open-source it (@juroosawa / The Information) (Visit Techmeme dot com for the link and full context!

@zephyr_z9

Big boi from Minimax Fable territory

Preview built on a synthetic news corpus (16 weeks, Apr–Jul 2026). Impact calls are model reads, not price data.

About Archive