Weight-only INT4 for RTX 3090/4090-class cards, where FP8 and NVFP4 are emulated or unusable. Full recipe on every card.
Alex Αdamopoulos
aleada
·
AI & ML interests
Verified open-weight inference for consumer GPUs. W4A16 packs, reproducible benchmarks and tools that inspect what quantized models actually contain.
Recent Activity
updated a model about 23 hours ago
aleada/Llama-3.2-11B-Vision-Instruct-W4A16 updated a model about 23 hours ago
aleada/Pixtral-12B-W4A16 updated a model about 23 hours ago
aleada/SmolLM3-3B-W4A16Organizations
None yet
spaces 6
Running
Find a model that fits
🔎
Search HF models by real size, quantization and topology
Running
Serving a model on your own GPU
🧭
Five questions between a model page and a model that runs
Running
Does this model need a reasoning parser?
🧩
Decide --reasoning-parser from a model's own chat template
Running
Does this pack keep what it claims?
🔎
Check a published quantized model against its own config
Running
How much of this pack is actually 4-bit?
🔍
What is really 4-bit in a quantized pack
Running
What context will actually fit?
📏
Real KV cache size from a model's declared attention
models 10
aleada/Llama-3.2-11B-Vision-Instruct-W4A16
Image-Text-to-Text • 11B • Updated • 68
aleada/Pixtral-12B-W4A16
Image-Text-to-Text • 13B • Updated • 99 • 1
aleada/SmolLM3-3B-W4A16
Text Generation • 3B • Updated • 729
aleada/Phi-4-mini-instruct-W4A16
Text Generation • 4B • Updated • 751
aleada/Phi-4-W4A16
Text Generation • 15B • Updated • 594
aleada/Gemma-3-12B-it-W4A16
Image-Text-to-Text • 12B • Updated • 40
aleada/Qwen3.8-27B-W4A16
Image-Text-to-Text • 28B • Updated • 71
aleada/Qwen3.6-27B-W4A16
Image-Text-to-Text • 28B • Updated • 161
aleada/Nemotron-3.5-Lightning-30B-A3B-W4A16
Text Generation • 33B • Updated • 374
aleada/Gemma-4-12B-it-W4A16
Image-Text-to-Text • 12B • Updated • 517
datasets 0
None public yet