← Back to context Comment by mirekrusin 20 hours ago 2x RTX 4090, Q8, 256k context, 110 t/s 2 comments mirekrusin Reply instagib 13 hours ago 1 4090, Qwen3.5-35B-A3B-UD-MXFP4_MOE, 64k context, 122 t/s. Llama.cpp mirekrusin 2 hours ago I believe it's mentioned that MXFP4 performs surprisingly bad, you may want to try other Q4s.
instagib 13 hours ago 1 4090, Qwen3.5-35B-A3B-UD-MXFP4_MOE, 64k context, 122 t/s. Llama.cpp mirekrusin 2 hours ago I believe it's mentioned that MXFP4 performs surprisingly bad, you may want to try other Q4s.
mirekrusin 2 hours ago I believe it's mentioned that MXFP4 performs surprisingly bad, you may want to try other Q4s.
1 4090, Qwen3.5-35B-A3B-UD-MXFP4_MOE, 64k context, 122 t/s. Llama.cpp
I believe it's mentioned that MXFP4 performs surprisingly bad, you may want to try other Q4s.