Omar Farouk
@omar-quantRegular
Quantization nerd. AMD GPUs and CPUs welcome.
- Quantization
- ROCm
- CPU inference
Cairo, Egypt · English, Arabic · Africa/Cairo
Pinned
quant-sweepsOct 8, 2026Outcome: Q2 to Q8 sweeps for 9 models, with the raw run records.Open sourceIndependently reproduced
- Establishes
- Another member re-ran it and matched the reported results.
- Doesn’t establish
- Does not establish results on other hardware.
- Disclosure
- Independent work
Runs
11 verified
ModelWhereSpeed
Qwen3.5-2B Q4_K_MNov 1, 2026RX 7900 XTX · 24 GB212.5 tok/s
Llama-3.1-8B-Instruct Q4_K_MOct 27, 2026AMD Ryzen 9 7950X · 64 GB9.5 tok/s
gemma-3-4b-it Q4_K_MOct 18, 2026RX 7900 XTX · 24 GB201.6 tok/s
Mistral-7B-Instruct-v0.2 Q4_K_MOct 10, 2026AMD Ryzen 9 7950X · 64 GB11.1 tok/s- gemma-3-12b-it Q4_K_MOct 5, 2026RX 7900 XTX · 24 GB66.6 tok/s
MiniCPM5-2B Q4_K_MSep 26, 2026AMD Ryzen 9 7950X · 64 GB31.5 tok/s- Qwen3-4B Q4_K_MSep 18, 2026RX 7900 XTX · 24 GB221.1 tok/s
- Qwen3.5-2B Q4_K_MSep 10, 2026AMD Ryzen 9 7950X · 64 GB32.0 tok/s
PassedFailedNot supported
Accounts, metrics and works carry separate evidence. Updated Nov 2, 2026. Public JSON