Team Ai
Modelpublic

renezander030/browserground-gguf

sourceHugging Faceapache-2.0updated 5mo agoView on Hugging Face
0likes34downloads
Model Card

browserground-gguf (llama.cpp / Ollama)

GGUF build of renezander030/browserground for llama.cpp, Ollama, and downstream wrappers that accept GGUF multimodal models.

Two files, both required:

FilePurposeSize
browserground-Q4_K_M.gguftext LLM, Q4KM quant1.11 GB
browserground-mmproj-f16.ggufvision tower (mmproj), f160.82 GB

Use via Ollama

A ready-made Modelfile is in the repo. After downloading both .gguf files:

bash
ollama create browserground -f Modelfile
ollama run browserground "Locate the Submit button" /path/to/screenshot.png

Use via llama.cpp directly

bash
llama-mtmd-cli   -m browserground-Q4_K_M.gguf   --mmproj browserground-mmproj-f16.gguf   --image screenshot.png   -p "Locate the element described: Submit button"

Or via the npm CLI (auto-routes to MLX on Apple Silicon)

bash
npm install -g browserground
browserground parse screenshot.png --target "Submit button"

Recipe, numbers, full evaluation: <https://huggingface.co/renezander030/browserground>.

License: Apache 2.0 (inherits from Qwen/Qwen3-VL-2B-Instruct).