Model reference · open weights

UniAR-RL

UniAR-RL is an open-weight language model from ShareLab-SII. UniAR-RL (BF16) weighs 19.3 GB; the smallest configuration that runs it is RTX 4090 24 GB.

For self-hosting, compare this build's requirements below with the GPU you plan to rent.

• UniAR-RL is a 9.6B parameter unified autoregressive multimodal model developed by ShareLab-SII for image understanding, generation, and editing.
• It uses a Qwen3-8B backbone and a BSQ visual tokenizer to process inputs within a 32,768 token context window.
• The model is trained with reinforcement learning and is available under the Apache 2.0 license for English tasks.

Summary of the ShareLab-SII/UniAR-RL model card, 2026-10-01

Hugging Face ↗Paper ↗