| model                          |       size |     params | backend    | ngl | type_k | type_v |  fa |            test |                  t/s |
| ------------------------------ | ---------: | ---------: | ---------- | --: | -----: | -----: | --: | --------------: | -------------------: |
| minimax-m2 230B.A10B Q4_K - Medium | 131.10 GiB |   228.69 B | CUDA       | 999 |   q8_0 |   q8_0 |   1 |           pp512 |     1817.28 ± 610.16 |
| minimax-m2 230B.A10B Q4_K - Medium | 131.10 GiB |   228.69 B | CUDA       | 999 |   q8_0 |   q8_0 |   1 |          pp4096 |      3496.87 ± 10.43 |
| minimax-m2 230B.A10B Q4_K - Medium | 131.10 GiB |   228.69 B | CUDA       | 999 |   q8_0 |   q8_0 |   1 |           tg128 |        129.57 ± 0.20 |
| minimax-m2 230B.A10B Q4_K - Medium | 131.10 GiB |   228.69 B | CUDA       | 999 |   q8_0 |   q8_0 |   1 |  pp512 @ d16384 |       1872.97 ± 5.06 |
| minimax-m2 230B.A10B Q4_K - Medium | 131.10 GiB |   228.69 B | CUDA       | 999 |   q8_0 |   q8_0 |   1 | pp4096 @ d16384 |       2581.81 ± 6.53 |
| minimax-m2 230B.A10B Q4_K - Medium | 131.10 GiB |   228.69 B | CUDA       | 999 |   q8_0 |   q8_0 |   1 |  tg128 @ d16384 |         89.67 ± 0.23 |

build: d2462f8f7 (9590)
