Beholder-q4f16 / version.json
GetBeholder's picture
v5: MIN short-prompt build (q4f16) + clean README
2d91503 verified
Raw
History Blame Contribute Delete
497 Bytes
{
"version": "v5",
"modelId": "Beholder-v5-q4f16",
"modelUrl": "https://huggingface.co/GetBeholder/Beholder-q4f16",
"modelLib": "https://huggingface.co/GetBeholder/Beholder-q4f16/resolve/main/Beholder-q4f16-webgpu.wasm",
"ggufUrl": "https://huggingface.co/GetBeholder/Beholder-GGUF",
"approxDownloadMB": 430,
"vramRequiredMB": 1300,
"notes": "Short-prompt build: smaller prompt, faster, improved accuracy. Use near-greedy decoding (low temperature) for stable structured output."
}