actio-ui-7b-rlvr-GGUF

actio-ui-7b-rlvr-GGUF

The ActIO-UI-7B-RLVR model from Uniphore is a 7B-parameter vision-language model fine-tuned from Qwen/Qwen2.5-VL-7B-Instruct using supervised fine-tuning (SFT) followed by reinforcement learning with verifiable rewards (RLVR), specialized for solving GUI subtasks like element grounding, interaction planning, and execution in computer-use agents, web agents, and multimodal environments under the open-mdw license. It achieves state-of-the-art performance among open-source 7B models on the WARC-Bench benchmark, scoring 78.09% on synthetic dev data, 54.44% on real dev data, 72.13% total dev, and 29.17% on test—outperforming peers like UI-Tars-1.5 7B (39.66% dev total) and Qwen2.5-VL 7B (15.54%) while competing with closed-source leaders like Claude Sonnet 3.7 in trajectory-level success rates. Designed for image-to-text pipelines with Transformers library support, it excels in GUI navigation, grounding, and agentic tasks, enabling efficient deployment for real-world UI automation.

actio-ui-7b-rlvr [GGUF]

Quants Usage

(sorted by size, not necessarily quality. IQ-quants are often preferable over similar sized non-IQ quants)

Here is a handy graph by ikawrakow comparing some lower-quality quant types (lower is better):