# A1 imatrix GGUF Comparison

Generated for local planning. No files were uploaded or published by this pipeline.

| Variant | Size | Decode tok/s | Accuracy | Retention vs F16 | Mean KLD | Same top p | Recommended Hardware | Best For |
|---|---:|---:|---:|---:|---:|---:|---|---|
| A1-Q2_K-imatrix | 12.05 GiB | 23.86 | 87.50% | 97.67% | 0.128242 | 81.75% | 8-12 GB RAM | smallest generated file |
| A1-IQ3_M-imatrix | 14.38 GiB | 16.27 | 89.58% | 100.00% | 0.045312 | 92.06% | 8-12 GB RAM | tiny but usable |
| A1-Q3_K_M-imatrix | 15.61 GiB | 20.13 | 89.58% | 100.00% | 0.046980 | 88.09% | 12-16 GB RAM | low-memory balance |
| A1-Q4_K_M-imatrix | 19.71 GiB | 24.28 | 87.50% | 97.67% | 0.015182 | 93.65% | 16-24 GB RAM | default recommendation |
| A1-IQ4_XS-imatrix | 17.44 GiB | 21.11 | 87.50% | 97.67% | 0.020185 | 92.06% | 16-24 GB RAM | efficient compact 4-bit |