GGUF models made with the method ("Mixture of Quantizations") proposed by Waleed Ahmad. I also used Unsloth M3's imatrix for calibration.
More details and evaluation here: MiniMax M3 GGUF Quantization: From 852 GB to ~150 GB Without Breaking Accuracy
Avoid using the MoQ-2.5.
- Compute Sponsorship: Verda. I used 2 B300s for quantization and evaluation.