llama-cli -hf liodon-ai/CodeLlama-7b-hf-imatrix-GGUF:Q4_K_Mollama run hf.co/liodon-ai/CodeLlama-7b-hf-imatrix-GGUF:Q4_K_Mliodon-ai/CodeLlama-7b-hf-imatrix-GGUF and pick your quant.| Quant | Size | VRAM est. | Notes |
|---|---|---|---|
IQ2_M | 2.36 GB | ~3 GB | 2-bit, iMatrix — smallest usable |
IQ3_M | 3.11 GB | ~4 GB | 3-bit, iMatrix — great quality/size tradeoff |
IQ4_XS | 3.62 GB | ~4 GB | 4-bit extra-small, iMatrix |
Q4_K_M | 4.08 GB | ~5 GB | 4-bit, iMatrix-calibrated (recommended) |
Q5_K_M | 4.78 GB | ~6 GB | 5-bit, iMatrix-calibrated |
Q6_K | 5.53 GB | ~6 GB | 6-bit, iMatrix-calibrated, near-lossless |
Q8_0 | 7.16 GB | ~8 GB | 8-bit, essentially lossless |
Also see plain (non-iMatrix) quants:liodon-ai/CodeLlama-7b-hf-GGUF