Two weeks after MiniMax posted about their 2.7 model on X, and 12 days after the model appeared on HuggingFace, the open weights haven't materialized in a usable form. That gap - announcement to actual release - has become a recurring frustration in the local AI community.
Open-weight models are ones where the company releases the raw model files so anyone can download and run the AI on their own hardware, without paying for API access. When a company announces open-weight availability but then delays or restricts the actual download, it creates a trust problem. People plan around these releases. Developers build tooling, hobbyists provision hardware, researchers design experiments.
MiniMax is a Chinese AI lab that's been putting out competitive models. Their 2.7 model, based on what was shared at announcement, is a capable language model in the mid-size range. But a model sitting on HuggingFace without accessible weights is functionally the same as no release at all.
This pattern isn't unique to MiniMax. Several labs have announced open releases and then quietly restricted access, required approval forms, or simply left placeholder pages up while the actual weights stayed offline. The local AI community has a long memory for this kind of thing, and it affects how much trust future announcements get.
No official statement from MiniMax explaining the delay has surfaced as of April 5, 2026.