gemma-4-12b-it
141 records across 16 hardware.
| hardware | model | quant | backend | tok/s | pp tok/s | W | ttft s | ctx/tg | date | class | flags | source |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| M5 Max | gemma-4-12b-it | 8bit | oMLX | 72.9 | 1721.0 | – | – | 1024 | 2026-10-17 | community | – | source |
| M5 Max | gemma-4-12b-it | 4bit | oMLX | 63.8 | 1176.0 | – | – | 1024 | 2026-10-16 | community | – | source |
| M5 Max | gemma-4-12b-it | 4bit | oMLX | 63.8 | 1176.0 | – | – | 1024 | 2026-10-16 | community | – | source |
| M4 Pro | gemma-4-12b-it | oQ8e | oMLX | 56.9 | 279.4 | – | – | 4096 | 2026-10-18 | community | – | source |
| M4 Pro | gemma-4-12b-it | oQ8e | oMLX | 56.9 | 279.4 | – | – | 4096 | 2026-10-18 | community | – | source |
| M5 Max | gemma-4-12b-it | 4bit | oMLX | 56.1 | 1307.0 | – | – | 65536 | 2026-10-16 | community | – | source |
| M5 Max | gemma-4-12b-it | 4bit | oMLX | 56.1 | 1307.0 | – | – | 65536 | 2026-10-16 | community | – | source |
| M3 Max | gemma-4-12b-it | oQ6e | oMLX | 54.7 | 446.9 | – | – | 1024 | 2026-10-15 | community | – | source |
| M5 Max | gemma-4-12b-it | oQ5e | oMLX | 51.8 | 1534.0 | – | – | 1024 | 2026-10-15 | community | – | source |
| M5 Max | gemma-4-12b-it | oQ5e | oMLX | 51.1 | 1823.0 | – | – | 4096 | 2026-10-15 | community | – | source |
| M5 Max | gemma-4-12b-it | 4bit | oMLX | 49.3 | 1023.0 | – | – | 131072 | 2026-10-16 | community | – | source |
| M5 Max | gemma-4-12b-it | 4bit | oMLX | 49.3 | 1023.0 | – | – | 131072 | 2026-10-16 | community | – | source |
| M5 Max | gemma-4-12b-it | 4bit | oMLX | 48.0 | 919.6 | – | – | 1024 | 2026-10-07 | community | – | source |
| M5 Max | gemma-4-12b-it | 4bit | oMLX | 47.8 | 1338.0 | – | – | 4096 | 2026-10-09 | community | – | source |
| M3 Max | gemma-4-12b-it | oQ6e | oMLX | 46.8 | 446.3 | – | – | 4096 | 2026-10-15 | community | – | source |
| M5 Max | gemma-4-12b-it | 4bit | oMLX | 46.5 | 1253.0 | – | – | 4096 | 2026-10-07 | community | – | source |
| M4 Pro | gemma-4-12b-it | oQ8e | oMLX | 46.4 | 261.9 | – | – | 16384 | 2026-10-18 | community | – | source |
| M4 Pro | gemma-4-12b-it | oQ8e | oMLX | 46.4 | 261.9 | – | – | 16384 | 2026-10-18 | community | – | source |
| M5 Max | gemma-4-12b-it | oQ5e | oMLX | 46.2 | 1234.0 | – | – | 65536 | 2026-10-15 | community | – | source |
| M5 Max | gemma-4-12b-it | 6bit | oMLX | 46.0 | 1783.0 | – | – | 4096 | 2026-10-10 | community | – | source |
| M5 Max | gemma-4-12b-it | 6bit | oMLX | 46.0 | 1795.0 | – | – | 4096 | 2026-10-10 | community | – | source |
| M5 Max | gemma-4-12b-it | 6bit | oMLX | 46.0 | 1787.0 | – | – | 4096 | 2026-10-10 | community | – | source |
| M5 Max | gemma-4-12b-it | 6bit | oMLX | 45.9 | 1495.0 | – | – | 4096 | 2026-10-10 | community | – | source |
| M5 Max | gemma-4-12b-it | 4bit | oMLX | 43.5 | 821.3 | – | – | 199680 | 2026-10-16 | community | – | source |
| M5 Max | gemma-4-12b-it | 4bit | oMLX | 43.5 | 821.3 | – | – | 199680 | 2026-10-16 | community | – | source |
| Radeon RX 6700 XT | gemma-4-12b-it 12B | IQ4_NL | llama.cpp (Vulkan) | 40.92 | 354.4 | – | – | – | 2026-08 | community | – | source |
| M4 Pro | gemma-4-12b-it | oQ8e | oMLX | 40.9 | 277.2 | – | – | 1024 | 2026-10-18 | community | – | source |
| M5 Max | gemma-4-12b-it | 8bit | oMLX | 38.6 | 1462.0 | – | – | 4096 | 2026-10-14 | community | – | source |
| M5 Max | gemma-4-12b-it | 8bit | oMLX | 38.2 | 1742.0 | – | – | 16384 | 2026-10-14 | community | – | source |
| M4 Pro | gemma-4-12b-it | oQ8e | oMLX | 37.2 | 280.9 | – | – | 1024 | 2026-10-18 | community | – | source |
| M4 Pro | gemma-4-12b-it | oQ8e | oMLX | 37.2 | 280.9 | – | – | 1024 | 2026-10-18 | community | – | source |
| M3 Max | gemma-4-12b-it | 4bit | oMLX | 35.8 | 400.4 | – | – | 4096 | 2026-10-11 | community | – | source |
| M3 Max | gemma-4-12b-it | 4bit | oMLX | 35.8 | 400.4 | – | – | 4096 | 2026-10-11 | community | – | source |
| M2 Max | gemma-4-12b-it | 4bit | oMLX | 35.4 | 245.5 | – | – | 4096 | 2026-10-13 | community | – | source |
| M5 Pro | gemma-4-12b-it | 4bit | oMLX | 34.9 | 858.2 | – | – | 4096 | 2026-10-15 | community | – | source |
| Radeon RX 6700 XT | gemma-4-12b-it 12B | IQ4_NL | llama.cpp (ROCm) | 34.6 | 653.9 | – | – | – | 2026-08 | community | – | source |
| M4 Max | gemma-4-12b-it | oQ8e | oMLX | 34.6 | 519.7 | – | – | 8192 | 2026-10-16 | community | – | source |
| M5 Pro | gemma-4-12b-it | 4bit | oMLX | 34.2 | 804.3 | – | – | 16384 | 2026-10-15 | community | – | source |
| M3 Max | gemma-4-12b-it | 4bit | oMLX | 34.0 | 383.4 | – | – | 1024 | 2026-10-11 | community | – | source |
| M3 Max | gemma-4-12b-it | 4bit | oMLX | 34.0 | 383.4 | – | – | 1024 | 2026-10-11 | community | – | source |
| M2 Max | gemma-4-12b-it | 4bit | oMLX | 34.0 | 234.7 | – | – | 16384 | 2026-10-13 | community | – | source |
| Radeon AI PRO R9700 | gemma-4-12b-it | FP8 | vLLM | 33.38 | – | – | – | – | 2026-10-06 | community | – | source |
| M5 Pro | gemma-4-12b-it | 4bit | oMLX | 33.2 | 729.0 | – | – | 32768 | 2026-10-15 | community | – | source |
| M5 Max | gemma-4-12b-it | 8bit | oMLX | 33.1 | 1787.0 | – | – | 4096 | 2026-10-17 | community | contradiction | source |
| M4 Pro | gemma-4-12b-it | oQ8e | oMLX | 32.3 | 274.2 | – | – | 8192 | 2026-10-18 | community | – | source |
| M4 Pro | gemma-4-12b-it | oQ8e | oMLX | 32.3 | 274.2 | – | – | 8192 | 2026-10-18 | community | – | source |
| M4 Max | gemma-4-12b-it | oQ8e | oMLX | 32.2 | 439.9 | – | – | 32768 | 2026-10-16 | community | – | source |
| M4 Pro | gemma-4-12b-it | 4bit | oMLX | 31.5 | 224.5 | – | – | 4096 | 2026-10-08 | community | – | source |
| M5 Pro | gemma-4-12b-it | 4bit | oMLX | 30.6 | 619.9 | – | – | 65536 | 2026-10-15 | community | – | source |
| M4 Pro | gemma-4-12b-it | 4bit | oMLX | 30.3 | 213.1 | – | – | 16384 | 2026-10-08 | community | – | source |
| M1 Max | gemma-4-12b-it | 4bit | oMLX | 28.6 | 230.5 | – | – | 1024 | 2026-10-12 | community | – | source |
| M4 Pro | gemma-4-12b-it | oQ8e | oMLX | 28.3 | 278.7 | – | – | 4096 | 2026-10-18 | community | contradiction | source |
| M4 Max | gemma-4-12b-it | oQ8e | oMLX | 27.5 | 311.6 | – | – | 131072 | 2026-10-16 | community | – | source |
| M3 Pro | gemma-4-12b-it | oQ4e | oMLX | 26.6 | 207.4 | – | – | 4096 | 2026-10-15 | community | – | source |
| M3 Max | gemma-4-12b-it | oQ6e | oMLX | 26.4 | 403.7 | – | – | 8192 | 2026-10-15 | community | – | source |
| M5 Pro | gemma-4-12b-it | oQ4e | oMLX | 26.2 | 686.0 | – | – | 1024 | 2026-10-17 | community | – | source |
| M3 Max | gemma-4-12b-it | oQ6e | oMLX | 25.3 | 383.4 | – | – | 16384 | 2026-10-15 | community | – | source |
| M4 | gemma-4-12b-it | oQ4e | oMLX | 24.6 | 143.1 | – | – | 1024 | 2026-10-12 | community | – | source |
| M4 | gemma-4-12b-it | 4bit | oMLX | 24.6 | 139.0 | – | – | 4096 | 2026-10-15 | community | – | source |
| M1 Max | gemma-4-12b-it | 4bit | oMLX | 24.5 | 189.3 | – | – | 4096 | 2026-10-12 | community | – | source |
| M4 | gemma-4-12b-it | 4bit | oMLX | 24.3 | 143.5 | – | – | 1024 | 2026-10-15 | community | – | source |
| M4 Pro | gemma-4-12b-it | 6bit | oMLX | 22.7 | 270.7 | – | – | 1024 | 2026-10-09 | community | – | source |
| M3 Max | gemma-4-12b-it | oQ6e | oMLX | 22.5 | 383.7 | – | – | 32768 | 2026-10-15 | community | – | source |
| M5 Max | gemma-4-12b-it | unknown | oMLX | 21.4 | 669.1 | – | – | 1024 | 2026-10-15 | community | – | source |
| M5 Max | gemma-4-12b-it | unknown | oMLX | 21.2 | 1610.0 | – | – | 4096 | 2026-10-15 | community | – | source |
| M3 Max | gemma-4-12b-it | oQ6e | oMLX | 21.1 | 352.4 | – | – | 65536 | 2026-10-15 | community | – | source |
| M4 | gemma-4-12b-it | 4bit | oMLX | 21.0 | 141.5 | – | – | 1024 | 2026-10-13 | community | contradiction | source |
| M4 | gemma-4-12b-it | oQ4e | oMLX | 20.7 | 121.5 | – | – | 1024 | 2026-10-09 | community | contradiction | source |
| M4 | gemma-4-12b-it | 4bit | oMLX | 20.5 | 138.0 | – | – | 4096 | 2026-10-13 | community | contradiction | source |
| M5 Max | gemma-4-12b-it | unknown | oMLX | 20.2 | 1354.0 | – | – | 65536 | 2026-10-15 | community | – | source |
| M6 | gemma-4-12b-it | 4bit | oMLX | 19.8 | 594.8 | – | – | 4096 | 2026-10-09 | community | – | source |
| M5 Pro | gemma-4-12b-it | MXFP8 | oMLX | 19.8 | 833.3 | – | – | 4096 | 2026-10-18 | community | – | source |
| M5 Pro | gemma-4-12b-it | MXFP8 | oMLX | 19.8 | 940.7 | – | – | 8192 | 2026-10-18 | community | – | source |
| M5 Pro | gemma-4-12b-it | MXFP8 | oMLX | 19.6 | 610.9 | – | – | 1024 | 2026-10-18 | community | – | source |
| M5 Pro | gemma-4-12b-it | MXFP8 | oMLX | 19.6 | 827.4 | – | – | 16384 | 2026-10-18 | community | – | source |
| M6 | gemma-4-12b-it | 4bit | oMLX | 19.4 | 604.2 | – | – | 16384 | 2026-10-09 | community | – | source |
| M1 Pro | gemma-4-12b-it | oQ4e | oMLX | 19.4 | 115.1 | – | – | 16384 | 2026-10-16 | community | – | source |
| M6 | gemma-4-12b-it | 4bit | oMLX | 19.0 | 558.3 | – | – | 1024 | 2026-10-13 | community | – | source |
| M6 | gemma-4-12b-it | 4bit | oMLX | 18.8 | 609.7 | – | – | 4096 | 2026-10-13 | community | – | source |
| M6 | gemma-4-12b-it | 4bit | oMLX | 18.1 | 567.3 | – | – | 4096 | 2026-10-13 | community | – | source |
| M3 Pro | gemma-4-12b-it | 4bit | oMLX | 18.1 | 217.0 | – | – | 1024 | 2026-10-14 | community | – | source |
| M1 Pro | gemma-4-12b-it | oQ4e | oMLX | 18.1 | 115.3 | – | – | 16384 | 2026-10-16 | community | – | source |
| M2 Pro | gemma-4-12b-it | 4bit | oMLX | 18.0 | 135.8 | – | – | 1024 | 2026-10-14 | community | – | source |
| M3 Pro | gemma-4-12b-it | 4bit | oMLX | 17.9 | 217.0 | – | – | 4096 | 2026-10-14 | community | – | source |
| M2 Pro | gemma-4-12b-it | 4bit | oMLX | 17.7 | 133.5 | – | – | 4096 | 2026-10-14 | community | – | source |
| M3 Pro | gemma-4-12b-it | 4bit | oMLX | 17.3 | 221.1 | – | – | 1024 | 2026-10-14 | community | – | source |
| M3 Pro | gemma-4-12b-it | oQ4e | oMLX | 17.2 | 210.8 | – | – | 1024 | 2026-10-15 | community | – | source |
| M4 Pro | gemma-4-12b-it | oQ8e | oMLX | 16.9 | 246.6 | – | – | 32768 | 2026-10-18 | community | – | source |
| M4 Pro | gemma-4-12b-it | oQ8e | oMLX | 16.9 | 246.6 | – | – | 32768 | 2026-10-18 | community | – | source |
| M2 Pro | gemma-4-12b-it | oQ4e | oMLX | 16.7 | 130.6 | – | – | 4096 | 2026-10-15 | community | – | source |
| M2 Pro | gemma-4-12b-it | oQ4e | oMLX | 16.7 | 130.6 | – | – | 4096 | 2026-10-15 | community | – | source |
| M5 | gemma-4-12b-it | 4bit | oMLX | 16.5 | 503.5 | – | – | 1024 | 2026-10-06 | community | – | source |
| M5 | gemma-4-12b-it | 4bit | oMLX | 16.5 | 506.8 | – | – | 4096 | 2026-10-06 | community | – | source |
| M5 | gemma-4-12b-it | 4bit | oMLX | 16.5 | 503.5 | – | – | 1024 | 2026-10-06 | community | – | source |
| M5 | gemma-4-12b-it | 4bit | oMLX | 16.5 | 506.8 | – | – | 4096 | 2026-10-06 | community | – | source |
| M4 Pro | gemma-4-12b-it | oQ8e | oMLX | 16.2 | 217.4 | – | – | 65536 | 2026-10-18 | community | – | source |
| M4 Pro | gemma-4-12b-it | oQ8e | oMLX | 16.2 | 217.4 | – | – | 65536 | 2026-10-18 | community | – | source |
| M4 | gemma-4-12b-it | oQ4e | oMLX | 15.8 | 123.2 | – | – | 16384 | 2026-10-18 | community | – | source |
| M2 Pro | gemma-4-12b-it | oQ4e | oMLX | 15.5 | 130.8 | – | – | 1024 | 2026-10-15 | community | – | source |
| M2 Pro | gemma-4-12b-it | oQ4e | oMLX | 15.5 | 130.8 | – | – | 1024 | 2026-10-15 | community | – | source |
| M4 | gemma-4-12b-it | oQ4e | oMLX | 15.4 | 139.2 | – | – | 4096 | 2026-10-12 | community | – | source |
| M6 | gemma-4-12b-it | 4bit | oMLX | 14.9 | 534.7 | – | – | 16384 | 2026-10-13 | community | contradiction | source |
| M3 Max | gemma-4-12b-it | oQ6e | oMLX | 14.9 | 275.5 | – | – | 131072 | 2026-10-15 | community | – | source |
| M3 Pro | gemma-4-12b-it | oQ8e | oMLX | 14.3 | 212.9 | – | – | 4096 | 2026-10-15 | community | – | source |
| M4 Pro | gemma-4-12b-it | oQ8e | oMLX | 13.8 | 175.1 | – | – | 131072 | 2026-10-18 | community | – | source |
| M4 Pro | gemma-4-12b-it | oQ8e | oMLX | 13.8 | 175.1 | – | – | 131072 | 2026-10-18 | community | – | source |
| M4 Pro | gemma-4-12b-it | oQ8e | oMLX | 13.7 | 174.9 | – | – | 131072 | 2026-10-18 | community | – | source |
| M4 | gemma-4-12b-it | oQ4e | oMLX | 13.2 | 135.2 | – | – | 4096 | 2026-10-18 | community | contradiction | source |
| M3 Pro | gemma-4-12b-it | oQ8e | oMLX | 12.9 | 214.0 | – | – | 1024 | 2026-10-15 | community | – | source |
| M4 | gemma-4-12b-it | oQ4e | oMLX | 12.7 | 122.1 | – | – | 16384 | 2026-10-18 | community | contradiction | source |
| M4 | gemma-4-12b-it | 4bit | oMLX | 12.5 | 141.9 | – | – | 1024 | 2026-10-12 | community | contradiction | source |
| M4 | gemma-4-12b-it | oQ4e | oMLX | 12.5 | 137.1 | – | – | 4096 | 2026-10-18 | community | contradiction | source |
| M4 | gemma-4-12b-it | 4bit | oMLX | 12.4 | 138.8 | – | – | 4096 | 2026-10-12 | community | contradiction | source |
| M3 Pro | gemma-4-12b-it | oQ4e | oMLX | 11.6 | 196.5 | – | – | 16384 | 2026-10-15 | community | – | source |
| M4 | gemma-4-12b-it | oQ4e | oMLX | 11.5 | 116.3 | – | – | 4096 | 2026-10-09 | community | contradiction | source |
| M3 Pro | gemma-4-12b-it | oQ8e | oMLX | 9.8 | 196.6 | – | – | 16384 | 2026-10-15 | community | – | source |
| M5 | gemma-4-12b-it | 8bit | oMLX | 9.3 | 369.8 | – | – | 1024 | 2026-10-15 | community | – | source |
| M5 | gemma-4-12b-it | 8bit | oMLX | 9.3 | 369.8 | – | – | 1024 | 2026-10-15 | community | – | source |
| M5 | gemma-4-12b-it | 8bit | oMLX | 8.8 | 436.0 | – | – | 16384 | 2026-10-12 | community | – | source |
| M5 | gemma-4-12b-it | 8bit | oMLX | 8.8 | 428.9 | – | – | 4096 | 2026-10-15 | community | – | source |
| M5 | gemma-4-12b-it | 8bit | oMLX | 8.8 | 428.9 | – | – | 4096 | 2026-10-15 | community | – | source |
| A100-SXM4-80GB | gemma-4-12b-it | BF16 | vLLM | – | 8268.4 | – | – | 2048 | 2026-10-13 | community | – | source |
| A100-SXM4-80GB | gemma-4-12b-it | BF16 | vLLM | – | 8906.6 | – | – | 2048 | 2026-10-13 | community | – | source |
| A100-SXM4-80GB | gemma-4-12b-it | BF16 | vLLM | – | 9141.9 | – | – | 2048 | 2026-10-13 | community | – | source |
| A100-SXM4-80GB | gemma-4-12b-it | BF16 | vLLM | – | 7900.1 | – | – | 4096 | 2026-10-13 | community | – | source |
| A100-SXM4-80GB | gemma-4-12b-it | BF16 | vLLM | – | 8435.7 | – | – | 4096 | 2026-10-13 | community | – | source |
| Radeon AI PRO R9700 | gemma-4-12b-it | FP8 | vLLM | – | 1102 | – | – | – | 2026-10-06 | community | – | source |
| A100-SXM4-80GB | gemma-4-12b-it | BF16 | vLLM | – | 8271.5 | – | – | 1024 | 2026-10 | community | – | source |
| A100-SXM4-80GB | gemma-4-12b-it | BF16 | vLLM | – | 9456.8 | – | – | 1024 | 2026-10 | community | – | source |
| A100-SXM4-80GB | gemma-4-12b-it | BF16 | vLLM | – | 9922.0 | – | – | 1024 | 2026-10 | community | – | source |
| A100-SXM4-80GB | gemma-4-12b-it | BF16 | vLLM | – | 10004.8 | – | – | 1024 | 2026-10 | community | – | source |
| A100-SXM4-80GB | gemma-4-12b-it | BF16 | vLLM | – | 4339.0 | – | – | 512 | 2026-10 | community | – | source |
| A100-SXM4-80GB | gemma-4-12b-it | BF16 | vLLM | – | 7754.7 | – | – | 512 | 2026-10 | community | – | source |
| A100-SXM4-80GB | gemma-4-12b-it | BF16 | vLLM | – | 8859.6 | – | – | 512 | 2026-10 | community | – | source |
| A100-SXM4-80GB | gemma-4-12b-it | BF16 | vLLM | – | 9818.5 | – | – | 512 | 2026-10 | community | – | source |
| A100-SXM4-80GB | gemma-4-12b-it | BF16 | vLLM | – | 893.6 | – | – | 128 | 2026-10 | community | – | source |
| A100-SXM4-80GB | gemma-4-12b-it | BF16 | vLLM | – | 2914.4 | – | – | 128 | 2026-10 | community | – | source |
| A100-SXM4-80GB | gemma-4-12b-it | BF16 | vLLM | – | 4529.7 | – | – | 128 | 2026-10 | community | – | source |
| A100-SXM4-80GB | gemma-4-12b-it | BF16 | vLLM | – | 6409.1 | – | – | 128 | 2026-10 | community | – | source |
| A100-SXM4-80GB | gemma-4-12b-it | BF16 | vLLM | – | 8137.0 | – | – | 4096 | 2026-10 | community | – | source |
| A100-SXM4-80GB | gemma-4-12b-it | BF16 | vLLM | – | 8549.3 | – | – | 4096 | 2026-10 | community | – | source |