W4A16 and FP8 quantizations of open models, calibrated on German text with static fp8 KV-cache scales baked in.
-
SinclairSchneider/Apertus-v1.5-8B-W4A16-KV8-de
Text Generation • 9B • Updated • 565 -
SinclairSchneider/Apertus-v1.5-70B-FP8-KV8
Text Generation • 72B • Updated • 28 -
SinclairSchneider/Apertus-v1.5-70B-W4A16-KV8
Text Generation • 13B • Updated • 151 -
SinclairSchneider/Mistral-Medium-3.5-128B-W4A16-KV8
Text Generation • 22B • Updated • 547
