EXL3 quantization of MN-12B-Mag-Mell-R1, 4 bits per weight.
HumanEval (argmax)
| Model | Q4 | Q6 | Q8 | FP16 |
|---|---|---|---|---|
MN-12B-Mag-Mell-R1-exl3-4bpw (mistral) |
72.6 | 71.3 | 73.2 | 72.0 |
MN-12B-Mag-Mell-R1-exl3-4bpw (chatml) |
71.3 | 73.2 | 73.2 | 73.8 |
MN-12B-Mag-Mell-R1-exl3-6bpw (mistral) |
74.4 | 74.4 | 74.4 | 73.8 |
MN-12B-Mag-Mell-R1-exl3-6bpw (chatml) |
76.8 | 72.0 | 72.0 | 71.3 |
Mistral-Nemo-Instruct-2407-exl3-4bpw (mistral) |
74.4 | 72.6 | 73.2 | 72.0 |
Mistral-Nemo-Instruct-2407-exl3-4bpw (chatml) |
70.1 | 72.0 | 71.3 | 72.6 |
Mistral-Nemo-Instruct-2407-exl3-6bpw (mistral) |
70.7 | 69.5 | 69.5 | 68.9 |
Mistral-Nemo-Instruct-2407-exl3-6bpw (chatml) |
68.3 | 70.1 | 69.5 | 68.9 |
Muse-12B-exl3-6bpw (mistral) |
54.9 | 54.3 | 54.9 | 52.4 |
Muse-12B-exl3-6bpw (chatml) |
54.9 | 55.5 | 54.3 | 54.9 |
- Downloads last month
- 8
Inference Providers
NEW
This model isn't deployed by any Inference Provider.
🙋
Ask for provider support
Model tree for isogen/MN-12B-Mag-Mell-R1-exl3-4bpw
Base model
inflatebot/MN-12B-Mag-Mell-R1