Nex-AGI Open Models: Model List, Licenses and Articles

September 26, 2026September 27, 2026

About Nex-AGI

Nex-AGI is an alliance initiated by the Shanghai Innovation Institute in China. Its nex-agi organization on Hugging Face says it “aims to build a sustainable, agency-driven, closed-loop, open-source ecosystem".

Nex-AGI stands out for building on foundation models released by others and post-training them specifically for agent use. According to the model card, Nex-N2 is post-trained on the Qwen3.5 series, with Pro built on Qwen3.5-397B-A17B and mini on Qwen3.5-35B-A3B-Base. The earlier Nex-N1 models were built on DeepSeek-V3.1, Qwen3 and InternLM3.

What It Releases

• Nex-N2.5 (September 2026): agent models for long-horizon tasks in real-world environments, such as operating computers and browsers and running and testing programs. It comes in three sizes, mini, Pro and Max. mini and Pro continue the multimodal (image-reading) foundations of Nex-N2, while Max is built on a 1.6T text-only MoE foundation model, described as Nex-AGI’s first complete post-training effort at trillion-parameter scale (our family page has a detailed explanation).
• Nex-N2 (June 2026): the Qwen3.5-based mini and Pro described above.
• Nex-N1 (November 2025 to January 2026): the previous generation, based on DeepSeek-V3.1, Qwen3 and InternLM3.

Licenses

All the Nex-N2.5, Nex-N2 and Nex-N1 models we checked are Apache-2.0. You may also need to follow the license of the base model they were built on. Read each model’s license text and the base model’s license before use. This summary is not legal advice.

Running Them Locally

• There are no official GGUF builds (as of 2026-09-26). Quantized builds come from third parties. Nex-N2 has official NVFP4 and FP8 builds.
• Only mini is a size you can run locally. mini is an MoE of about 35B total, and its config file has the same structure as Qwen3.5-35B-A3B, the base of Nex-N2-mini (8 of 256 experts active). With quantization, consumer GPUs are candidates. Pro (about 397B) and Max (about 1.6T) are data-center sizes.
• It is also available as a hosted service. The model card points to OpenRouter.

Sources: the nex-agi organization card; model cards of nex-agi/Nex-N2.5-Pro and nex-agi/Nex-N2-Pro; the model list of the nex-agi organization on Hugging Face (all as of 2026-09-26).

Our Coverage and Data

Local Model Watch has published 2 article(s) on Nex-AGI’s own models, 0 on third-party fine-tunes and quantized builds based on them, and 0 from its official blog. The model list below is collected by code from Hugging Face every day. Part of our publisher index; model families link to their family pages.

At a Glance

Recent Models on Hugging Face

The publisher lists 14 models in total on Hugging Face. Below are up to 15 of the most recent generative models (text, image, video, audio and so on), the scope of this site, per organization. Classifiers, feature extractors, research components and repositories without a model card are omitted, and the publisher’s own quantized builds and helper variants are folded into the row of the original model. Parameter counts are left blank for repositories that pack weights into integer types, where Hugging Face’s count is not the real parameter count.

From the Hugging Face API. Last changed 2026-09-26 (JST). Collected by code.

Licenses of Its Own Models (as recorded in our articles)

The license can differ from model to model even within one publisher. This is not legal advice; check each model’s license text before use.

By Model Family

Articles on Its Own Models

Last updated 2026-09-26 (JST). The explanation at the top of this page was written with the help of AI from the primary sources it cites. The tables and lists under “Our Coverage and Data" are assembled by code from our article log and the Hugging Face API.