Back to Model
N

NorthCUA-Nano

Model Nov 2, 2026

An independent dense 9B computer-use model built on Qwen3.5, with downloadable BF16 weights and native tool-call formatting for an external desktop or browser runtime.

Visit primary link 3 sources attached
NorthCUA project diagram connecting a desktop and web harness, experience collection and staged model training.
Project-published Figure 2 from the NorthCUA technical report. GUI states, cursor cues and rewards are schematic; the hard-task panel includes proposed collection methods. This illustrates the training pipeline, not an AITrending test or proof of performance for the released checkpoint. Source: NorthCUA project
self-hosted

What happened lately

NorthCUA-Nano timeline

NorthCUA-Nano is an independent Qwen3.5-9B post-training release for computer-use agents. The November 2 export supplies four predominantly BF16 Safetensors shards, configuration, tokenizer, image processor, native chat template and checksums under Apache 2.0. Its English card describes approximately 18.84 GB of model files before runtime caches.

An external runtime must execute tool calls and return updated observations. The card records Transformers 5.20.2 and provides an illustrative screenshot-generation example, not a tested serving configuration. Research benchmark results on the card are explicitly not tied to the exact checkpoint975 export; AITrending has not reproduced them.