Nex AGI releases Nex-N2.5-mini, an open MoE multimodal model

Nex AGI has published Nex-N2.5-mini, a compact multimodal model that combines text and vision capabilities in a mixture-of-experts (MoE) architecture. The model is available on Hugging Face under the permissive Apache-2.0 license, which allows both commercial and research use with minimal restrictions.

The "mini" designation signals that this is positioned as a smaller entry, likely aimed at developers who want capable image-and-text reasoning without the footprint of a flagship system. Its MoE design is a notable choice: by routing each input through only a subset of the model's parameters, MoE architectures can deliver strong performance while keeping the active compute per token lower than a dense model of comparable size.

Why it matters

Open multimodal models under truly permissive licenses remain relatively scarce, and each new release widens the pool of tools teams can build on freely.

• Handles both text and vision inputs
• Uses a mixture-of-experts architecture for efficiency
• Ships under Apache-2.0, enabling commercial use

Nex AGI has not published detailed specifications such as parameter count or context length alongside this initial release, so real-world performance will depend on independent testing. As the first entry in the N2.5 line, the model gives the community an early look at the company's approach to efficient, open multimodal AI.