W4A16 for consumer Ampere
Collection
Weight-only INT4 for RTX 3090/4090-class cards, where FP8 and NVFP4 are emulated or unusable. Full recipe on every card. • 12 items • Updated
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 collection 20 days ago
W4A16 for consumer Ampere updated a collection 20 days ago
W4A16 for consumer Ampere updated a model 20 days ago
aleada/Nemotron-3-Nano-4B-W4A16Organizations
None yet
updated a
collection 20 days ago
updated a
model 20 days ago
published a
model 20 days ago
updated a
model 22 days ago
updated a
collection 22 days ago
updated 10
models 23 days ago
aleada/SmolLM3-3B-W4A16
Text Generation • 3B • Updated • 315
aleada/Qwen3.8-27B-W4A16
Image-Text-to-Text • 28B • Updated • 174
aleada/Qwen3.6-27B-W4A16
Image-Text-to-Text • 28B • Updated • 46
aleada/Pixtral-12B-W4A16
Image-Text-to-Text • 13B • Updated • 1.22k • 1
aleada/Phi-4-mini-instruct-W4A16
Text Generation • 4B • Updated • 353
aleada/Phi-4-W4A16
Text Generation • 15B • Updated • 375
aleada/Nemotron-3.5-Lightning-30B-A3B-W4A16
Text Generation • 33B • Updated • 420
aleada/Llama-3.2-11B-Vision-Instruct-W4A16
Image-Text-to-Text • 11B • Updated • 33
aleada/Gemma-4-12B-it-W4A16
Image-Text-to-Text • 12B • Updated • 610
aleada/Gemma-3-12B-it-W4A16
Image-Text-to-Text • 12B • Updated • 87
published a
model 23 days ago
updated a
Space about 1 month ago
published a
Space about 1 month ago
updated a
Space about 1 month ago