{ "hardware": "Apple M5 Max, 128 GB unified memory (MacBook Pro)", "builds": { "gmlx": "0.1.0", "mlx-kquant": "0.3.5", "llama.cpp": "b9967", "ds4-server": { "deepseek-v4-flash-antirez-iq2xxs": "b030961", "deepseek-v4.1-flash-antirez-q2": "8db1d1d" } }, "order": [ "qwen3.5-122b-a10b-ud-q5km", "qwen3.6-35b-a3b-heretic-q6k", "qwen3.6-27b-q6k", "qwen3.5-9b-q6k", "gemma-4-31b-q6k", "gemma-4-26b-a4b-q6k", "gemma-4-12b-q6k", "gemma-4-e4b-q6k", "gemma-4-e2b-q6kxl", "gpt-oss-120b-heretic-mxfp4", "gpt-oss-20b-mxfp4", "dolphin3-llama3.1-8b-q6k", "deepseek-v4-flash-unsloth-udiq3xxs", "deepseek-v4-flash-antirez-iq2xxs", "deepseek-v4.1-flash-antirez-q2", "qwen38-flash-next-udq3kxl" ], "models": { "qwen3.5-122b-a10b-ud-q5km": { "label": "Qwen3.5-122B-A10B UD-Q5_K_M", "comparison": "llama.cpp", "mtp": true, "mtp_kind": "native", "gguf": "Qwen3.5-122B-A10B-UD-Q5_K_M-00001-of-00003.gguf", "source": "https://huggingface.co/unsloth/Qwen3.5-122B-A10B-MTP-GGUF", "builds": { "gmlx": "0.1.0", "mlx-kquant": "0.3.5", "llama.cpp": "b9967" }, "measured": "2026-07-18", "rows": [ { "depth": 512, "gmlx_decode_base": 47.8, "gmlx_decode_mtp": 59.0, "llama_decode_base": 38.4, "llama_decode_mtp": 47.8, "gmlx_prefill_base": 643.9, "llama_prefill_base": 568.8, "range": { "gmlx_decode_base": [ 47.4, 48.0 ], "gmlx_decode_mtp": [ 55.6, 67.3 ], "llama_decode_base": [ 38.0, 38.5 ], "llama_decode_mtp": [ 42.3, 51.0 ], "gmlx_prefill_base": [ 482.2, 706.7 ], "llama_prefill_base": [ 546.6, 612.4 ] } }, { "depth": 4300, "gmlx_decode_base": 45.9, "gmlx_decode_mtp": 58.3, "llama_decode_base": 37.3, "llama_decode_mtp": 46.4, "gmlx_prefill_base": 950.3, "llama_prefill_base": 555.0, "range": { "gmlx_decode_base": [ 45.7, 46.2 ], "gmlx_decode_mtp": [ 55.6, 62.8 ], "llama_decode_base": [ 37.1, 37.5 ], "llama_decode_mtp": [ 44.7, 48.0 ], "gmlx_prefill_base": [ 935.3, 956.4 ], "llama_prefill_base": [ 496.5, 607.6 ] } }, { "depth": 17000, "gmlx_decode_base": 42.2, "gmlx_decode_mtp": 56.0, "llama_decode_base": 33.8, "llama_decode_mtp": 38.6, "gmlx_prefill_base": 957.6, "llama_prefill_base": 476.3, "range": { "gmlx_decode_base": [ 42.1, 42.4 ], "gmlx_decode_mtp": [ 46.6, 61.2 ], "llama_decode_base": [ 33.7, 34.0 ], "llama_decode_mtp": [ 37.4, 40.7 ], "gmlx_prefill_base": [ 945.2, 968.5 ], "llama_prefill_base": [ 449.9, 491.7 ] } }, { "depth": 67000, "gmlx_decode_base": 33.6, "gmlx_decode_mtp": 42.0, "llama_decode_base": 25.6, "llama_decode_mtp": 25.4, "gmlx_prefill_base": 649.5, "llama_prefill_base": 284.8, "range": { "gmlx_decode_base": [ 32.8, 33.9 ], "gmlx_decode_mtp": [ 38.5, 44.7 ], "llama_decode_base": [ 25.4, 25.8 ], "llama_decode_mtp": [ 24.4, 31.2 ], "gmlx_prefill_base": [ 628.0, 654.5 ], "llama_prefill_base": [ 276.0, 289.7 ] } }, { "depth": 110000, "gmlx_decode_base": 28.4, "gmlx_decode_mtp": 36.7, "llama_decode_base": 21.1, "llama_decode_mtp": 21.9, "gmlx_prefill_base": 474.2, "llama_prefill_base": 202.2, "range": { "gmlx_decode_base": [ 27.9, 28.8 ], "gmlx_decode_mtp": [ 34.4, 38.2 ], "llama_decode_base": [ 21.0, 21.1 ], "llama_decode_mtp": [ 21.1, 24.7 ], "gmlx_prefill_base": [ 463.2, 478.4 ], "llama_prefill_base": [ 201.2, 207.3 ] } }, { "depth": 200000, "gmlx_decode_base": 21.9, "gmlx_decode_mtp": 22.0, "llama_decode_base": 15.6, "llama_decode_mtp": 14.4, "gmlx_prefill_base": 300.3, "llama_prefill_base": 129.9, "range": { "gmlx_decode_base": [ 21.8, 22.0 ], "gmlx_decode_mtp": [ 20.6, 23.8 ], "llama_decode_base": [ 15.5, 15.8 ], "llama_decode_mtp": [ 13.9, 15.9 ], "gmlx_prefill_base": [ 297.5, 301.3 ], "llama_prefill_base": [ 129.3, 130.0 ] } } ] }, "qwen3.6-35b-a3b-heretic-q6k": { "label": "Qwen3.6-35B-A3B Q6_K", "comparison": "llama.cpp", "mtp": true, "mtp_kind": "native", "gguf": "Qwen3.6-35B-A3B-uncensored-heretic-Native-MTP-Preserved-Q6_K.gguf", "source": "https://huggingface.co/llmfan46/Qwen3.6-35B-A3B-uncensored-heretic-Native-MTP-Preserved-GGUF", "builds": { "gmlx": "0.1.0", "mlx-kquant": "0.3.5", "llama.cpp": "b9967" }, "measured": "2026-07-18", "rows": [ { "depth": 512, "gmlx_decode_base": 128.8, "gmlx_decode_mtp": 138.1, "llama_decode_base": 89.3, "llama_decode_mtp": 110.1, "gmlx_prefill_base": 2106.6, "llama_prefill_base": 1737.9, "range": { "gmlx_decode_base": [ 126.7, 130.0 ], "gmlx_decode_mtp": [ 131.4, 143.1 ], "llama_decode_base": [ 88.2, 94.8 ], "llama_decode_mtp": [ 99.9, 115.3 ], "gmlx_prefill_base": [ 1554.9, 2319.9 ], "llama_prefill_base": [ 1166.7, 1756.3 ] } }, { "depth": 4300, "gmlx_decode_base": 128.1, "gmlx_decode_mtp": 139.0, "llama_decode_base": 94.4, "llama_decode_mtp": 104.0, "gmlx_prefill_base": 2764.5, "llama_prefill_base": 1722.3, "range": { "gmlx_decode_base": [ 127.5, 128.7 ], "gmlx_decode_mtp": [ 123.1, 151.3 ], "llama_decode_base": [ 93.2, 95.3 ], "llama_decode_mtp": [ 96.9, 111.1 ], "gmlx_prefill_base": [ 2636.6, 2849.5 ], "llama_prefill_base": [ 1478.8, 1823.7 ] } }, { "depth": 17000, "gmlx_decode_base": 114.3, "gmlx_decode_mtp": 118.8, "llama_decode_base": 85.7, "llama_decode_mtp": 87.4, "gmlx_prefill_base": 2736.4, "llama_prefill_base": 1440.5, "range": { "gmlx_decode_base": [ 113.9, 114.5 ], "gmlx_decode_mtp": [ 117.1, 130.5 ], "llama_decode_base": [ 85.2, 86.4 ], "llama_decode_mtp": [ 83.4, 103.3 ], "gmlx_prefill_base": [ 2685.6, 2761.8 ], "llama_prefill_base": [ 1403.8, 1459.5 ] } }, { "depth": 67000, "gmlx_decode_base": 83.4, "gmlx_decode_mtp": 85.0, "llama_decode_base": 58.4, "llama_decode_mtp": 63.2, "gmlx_prefill_base": 1841.4, "llama_prefill_base": 739.0, "range": { "gmlx_decode_base": [ 82.8, 84.7 ], "gmlx_decode_mtp": [ 80.7, 89.8 ], "llama_decode_base": [ 57.6, 59.9 ], "llama_decode_mtp": [ 52.4, 70.1 ], "gmlx_prefill_base": [ 1834.8, 1849.8 ], "llama_prefill_base": [ 727.1, 798.2 ] } }, { "depth": 110000, "gmlx_decode_base": 68.5, "gmlx_decode_mtp": 72.6, "llama_decode_base": 47.3, "llama_decode_mtp": 48.6, "gmlx_prefill_base": 1364.7, "llama_prefill_base": 537.6, "range": { "gmlx_decode_base": [ 68.4, 69.1 ], "gmlx_decode_mtp": [ 66.0, 80.3 ], "llama_decode_base": [ 46.9, 48.8 ], "llama_decode_mtp": [ 40.9, 50.8 ], "gmlx_prefill_base": [ 1338.3, 1374.4 ], "llama_prefill_base": [ 517.0, 561.6 ] } }, { "depth": 200000, "gmlx_decode_base": 46.8, "gmlx_decode_mtp": 48.3, "llama_decode_base": 33.9, "llama_decode_mtp": 31.3, "gmlx_prefill_base": 805.1, "llama_prefill_base": 321.1, "range": { "gmlx_decode_base": [ 46.5, 48.2 ], "gmlx_decode_mtp": [ 46.2, 50.7 ], "llama_decode_base": [ 33.6, 34.3 ], "llama_decode_mtp": [ 27.2, 36.3 ], "gmlx_prefill_base": [ 785.0, 835.0 ], "llama_prefill_base": [ 315.8, 324.3 ] } } ] }, "qwen3.6-27b-q6k": { "label": "Qwen3.6-27B Q6_K", "comparison": "llama.cpp", "mtp": true, "mtp_kind": "native", "gguf": "Qwen_Qwen3.6-27B-Q6_K.gguf", "source": "https://huggingface.co/bartowski/Qwen_Qwen3.6-27B-GGUF", "builds": { "gmlx": "0.1.0", "mlx-kquant": "0.3.5", "llama.cpp": "b9967" }, "measured": "2026-07-18", "rows": [ { "depth": 512, "gmlx_decode_base": 20.7, "gmlx_decode_mtp": 36.8, "llama_decode_base": 18.7, "llama_decode_mtp": 31.5, "gmlx_prefill_base": 430.2, "llama_prefill_base": 394.6, "range": { "gmlx_decode_base": [ 20.7, 20.8 ], "gmlx_decode_mtp": [ 35.1, 39.9 ], "llama_decode_base": [ 18.6, 19.2 ], "llama_decode_mtp": [ 31.1, 32.6 ], "gmlx_prefill_base": [ 407.4, 489.8 ], "llama_prefill_base": [ 335.7, 446.8 ] } }, { "depth": 4300, "gmlx_decode_base": 19.8, "gmlx_decode_mtp": 37.1, "llama_decode_base": 18.1, "llama_decode_mtp": 28.7, "gmlx_prefill_base": 600.9, "llama_prefill_base": 443.9, "range": { "gmlx_decode_base": [ 19.7, 20.1 ], "gmlx_decode_mtp": [ 33.9, 47.0 ], "llama_decode_base": [ 18.0, 18.3 ], "llama_decode_mtp": [ 27.5, 56.4 ], "gmlx_prefill_base": [ 596.0, 606.2 ], "llama_prefill_base": [ 412.5, 464.4 ] } }, { "depth": 17000, "gmlx_decode_base": 18.8, "gmlx_decode_mtp": 34.1, "llama_decode_base": 17.3, "llama_decode_mtp": 25.6, "gmlx_prefill_base": 575.9, "llama_prefill_base": 405.4, "range": { "gmlx_decode_base": [ 18.7, 18.8 ], "gmlx_decode_mtp": [ 29.5, 39.3 ], "llama_decode_base": [ 17.3, 17.4 ], "llama_decode_mtp": [ 24.1, 29.8 ], "gmlx_prefill_base": [ 570.1, 587.6 ], "llama_prefill_base": [ 399.7, 425.8 ] } }, { "depth": 67000, "gmlx_decode_base": 15.8, "gmlx_decode_mtp": 24.6, "llama_decode_base": 14.1, "llama_decode_mtp": 18.8, "gmlx_prefill_base": 454.5, "llama_prefill_base": 252.1, "range": { "gmlx_decode_base": [ 15.5, 16.2 ], "gmlx_decode_mtp": [ 21.8, 31.3 ], "llama_decode_base": [ 14.0, 14.1 ], "llama_decode_mtp": [ 17.6, 21.6 ], "gmlx_prefill_base": [ 436.6, 466.1 ], "llama_prefill_base": [ 247.9, 255.7 ] } }, { "depth": 110000, "gmlx_decode_base": 14.5, "gmlx_decode_mtp": 22.2, "llama_decode_base": 12.1, "llama_decode_mtp": 16.9, "gmlx_prefill_base": 387.7, "llama_prefill_base": 182.6, "range": { "gmlx_decode_base": [ 14.5, 14.6 ], "gmlx_decode_mtp": [ 18.5, 23.8 ], "llama_decode_base": [ 10.9, 12.2 ], "llama_decode_mtp": [ 13.1, 18.5 ], "gmlx_prefill_base": [ 384.6, 389.3 ], "llama_prefill_base": [ 160.3, 185.0 ] } }, { "depth": 200000, "gmlx_decode_base": 12.1, "gmlx_decode_mtp": 16.3, "llama_decode_base": 9.6, "llama_decode_mtp": 13.5, "gmlx_prefill_base": 280.3, "llama_prefill_base": 112.8, "range": { "gmlx_decode_base": [ 11.7, 12.4 ], "gmlx_decode_mtp": [ 15.3, 17.1 ], "llama_decode_base": [ 9.6, 9.6 ], "llama_decode_mtp": [ 11.2, 14.8 ], "gmlx_prefill_base": [ 273.1, 285.8 ], "llama_prefill_base": [ 112.5, 114.5 ] } } ] }, "qwen3.5-9b-q6k": { "label": "Qwen3.5-9B Q6_K", "comparison": "llama.cpp", "mtp": true, "mtp_kind": "native", "gguf": "Qwen3.5-9B-Q6_K.gguf", "source": "https://huggingface.co/unsloth/Qwen3.5-9B-MTP-GGUF", "builds": { "gmlx": "0.1.0", "mlx-kquant": "0.3.5", "llama.cpp": "b9967" }, "measured": "2026-07-18", "rows": [ { "depth": 512, "gmlx_decode_base": 70.3, "gmlx_decode_mtp": 111.9, "llama_decode_base": 58.1, "llama_decode_mtp": 75.8, "gmlx_prefill_base": 1595.0, "llama_prefill_base": 1136.4, "range": { "gmlx_decode_base": [ 69.1, 72.4 ], "gmlx_decode_mtp": [ 102.9, 126.7 ], "llama_decode_base": [ 57.5, 58.3 ], "llama_decode_mtp": [ 72.8, 80.9 ], "gmlx_prefill_base": [ 1407.5, 2267.2 ], "llama_prefill_base": [ 1075.4, 1142.4 ] } }, { "depth": 4300, "gmlx_decode_base": 69.1, "gmlx_decode_mtp": 96.6, "llama_decode_base": 58.2, "llama_decode_mtp": 76.8, "gmlx_prefill_base": 1817.4, "llama_prefill_base": 1424.0, "range": { "gmlx_decode_base": [ 69.0, 69.6 ], "gmlx_decode_mtp": [ 85.7, 103.0 ], "llama_decode_base": [ 58.0, 59.2 ], "llama_decode_mtp": [ 69.5, 86.7 ], "gmlx_prefill_base": [ 1786.1, 1916.4 ], "llama_prefill_base": [ 1342.0, 1582.0 ] } }, { "depth": 17000, "gmlx_decode_base": 62.7, "gmlx_decode_mtp": 85.1, "llama_decode_base": 53.4, "llama_decode_mtp": 68.4, "gmlx_prefill_base": 1962.4, "llama_prefill_base": 1350.5, "range": { "gmlx_decode_base": [ 62.0, 63.5 ], "gmlx_decode_mtp": [ 76.7, 91.9 ], "llama_decode_base": [ 53.3, 53.7 ], "llama_decode_mtp": [ 61.3, 73.6 ], "gmlx_prefill_base": [ 1920.0, 1997.6 ], "llama_prefill_base": [ 1306.0, 1375.9 ] } }, { "depth": 67000, "gmlx_decode_base": 49.8, "gmlx_decode_mtp": 58.5, "llama_decode_base": 39.6, "llama_decode_mtp": 49.4, "gmlx_prefill_base": 1551.7, "llama_prefill_base": 758.7, "range": { "gmlx_decode_base": [ 49.3, 50.7 ], "gmlx_decode_mtp": [ 52.5, 64.1 ], "llama_decode_base": [ 39.0, 40.8 ], "llama_decode_mtp": [ 43.9, 59.4 ], "gmlx_prefill_base": [ 1540.2, 1600.1 ], "llama_prefill_base": [ 740.7, 798.9 ] } } ] }, "gemma-4-31b-q6k": { "label": "gemma-4-31B-it Q6_K", "comparison": "llama.cpp", "mtp": true, "mtp_kind": "drafter", "gguf": "gemma-4-31B-it-Q6_K.gguf", "source": null, "builds": { "gmlx": "0.1.0", "mlx-kquant": "0.3.5", "llama.cpp": "b9967" }, "measured": "2026-07-18", "rows": [ { "depth": 512, "gmlx_decode_base": 17.8, "gmlx_decode_mtp": 34.5, "llama_decode_base": 16.4, "llama_decode_mtp": 28.4, "gmlx_prefill_base": 367.5, "llama_prefill_base": 281.1, "range": { "gmlx_decode_base": [ 17.6, 18.0 ], "gmlx_decode_mtp": [ 32.5, 37.5 ], "llama_decode_base": [ 16.4, 16.9 ], "llama_decode_mtp": [ 25.1, 28.8 ], "gmlx_prefill_base": [ 360.3, 384.2 ], "llama_prefill_base": [ 261.0, 283.8 ] } }, { "depth": 4300, "gmlx_decode_base": 16.1, "gmlx_decode_mtp": 30.0, "llama_decode_base": 15.6, "llama_decode_mtp": 24.8, "gmlx_prefill_base": 461.8, "llama_prefill_base": 341.7, "range": { "gmlx_decode_base": [ 15.2, 17.0 ], "gmlx_decode_mtp": [ 26.5, 31.7 ], "llama_decode_base": [ 15.5, 15.8 ], "llama_decode_mtp": [ 21.5, 27.1 ], "gmlx_prefill_base": [ 439.0, 490.4 ], "llama_prefill_base": [ 317.1, 354.5 ] } }, { "depth": 17000, "gmlx_decode_base": 15.2, "gmlx_decode_mtp": 29.1, "llama_decode_base": 14.5, "llama_decode_mtp": 21.5, "gmlx_prefill_base": 439.0, "llama_prefill_base": 292.4, "range": { "gmlx_decode_base": [ 13.9, 16.1 ], "gmlx_decode_mtp": [ 27.4, 34.4 ], "llama_decode_base": [ 14.4, 14.7 ], "llama_decode_mtp": [ 18.6, 25.6 ], "gmlx_prefill_base": [ 407.9, 463.6 ], "llama_prefill_base": [ 284.9, 295.8 ] } }, { "depth": 67000, "gmlx_decode_base": 12.4, "gmlx_decode_mtp": 23.6, "llama_decode_base": 11.1, "llama_decode_mtp": 14.1, "gmlx_prefill_base": 343.1, "llama_prefill_base": 166.0, "range": { "gmlx_decode_base": [ 11.1, 13.2 ], "gmlx_decode_mtp": [ 19.9, 25.5 ], "llama_decode_base": [ 11.1, 11.2 ], "llama_decode_mtp": [ 13.2, 18.2 ], "gmlx_prefill_base": [ 319.2, 363.7 ], "llama_prefill_base": [ 163.9, 166.7 ] } }, { "depth": 110000, "gmlx_decode_base": 11.3, "gmlx_decode_mtp": 18.2, "llama_decode_base": 9.5, "llama_decode_mtp": 10.3, "gmlx_prefill_base": 292.3, "llama_prefill_base": 118.7, "range": { "gmlx_decode_base": [ 11.2, 11.4 ], "gmlx_decode_mtp": [ 17.0, 19.6 ], "llama_decode_base": [ 9.5, 9.8 ], "llama_decode_mtp": [ 9.6, 13.5 ], "gmlx_prefill_base": [ 290.5, 293.3 ], "llama_prefill_base": [ 116.9, 122.8 ] } }, { "depth": 200000, "gmlx_decode_base": 9.0, "gmlx_decode_mtp": 12.4, "llama_decode_base": 7.3, "llama_decode_mtp": 6.4, "gmlx_prefill_base": 204.2, "llama_prefill_base": 74.1, "range": { "gmlx_decode_base": [ 9.0, 9.1 ], "gmlx_decode_mtp": [ 11.8, 13.5 ], "llama_decode_base": [ 6.9, 7.7 ], "llama_decode_mtp": [ 6.0, 8.2 ], "gmlx_prefill_base": [ 192.2, 207.7 ], "llama_prefill_base": [ 72.1, 75.4 ] } } ] }, "gemma-4-26b-a4b-q6k": { "label": "gemma-4-26B-A4B-it Q6_K", "comparison": "llama.cpp", "mtp": true, "mtp_kind": "drafter", "gguf": "google_gemma-4-26B-A4B-it-Q6_K.gguf", "source": "https://huggingface.co/bartowski/google_gemma-4-26B-A4B-it-GGUF", "builds": { "gmlx": "0.1.0", "mlx-kquant": "0.3.5", "llama.cpp": "b9967" }, "measured": "2026-07-18", "rows": [ { "depth": 512, "gmlx_decode_base": 97.8, "gmlx_decode_mtp": 113.9, "llama_decode_base": 82.6, "llama_decode_mtp": 93.0, "gmlx_prefill_base": 1821.3, "llama_prefill_base": 1268.3, "range": { "gmlx_decode_base": [ 96.9, 98.8 ], "gmlx_decode_mtp": [ 104.2, 127.1 ], "llama_decode_base": [ 80.0, 82.6 ], "llama_decode_mtp": [ 90.4, 93.5 ], "gmlx_prefill_base": [ 1531.2, 2125.0 ], "llama_prefill_base": [ 1104.3, 1480.1 ] } }, { "depth": 4300, "gmlx_decode_base": 97.2, "gmlx_decode_mtp": 107.1, "llama_decode_base": 85.5, "llama_decode_mtp": 96.3, "gmlx_prefill_base": 2610.2, "llama_prefill_base": 1549.2, "range": { "gmlx_decode_base": [ 97.1, 97.3 ], "gmlx_decode_mtp": [ 98.1, 117.7 ], "llama_decode_base": [ 81.9, 90.5 ], "llama_decode_mtp": [ 79.2, 109.9 ], "gmlx_prefill_base": [ 2514.4, 2668.2 ], "llama_prefill_base": [ 1400.0, 1650.7 ] } }, { "depth": 17000, "gmlx_decode_base": 90.4, "gmlx_decode_mtp": 95.9, "llama_decode_base": 75.9, "llama_decode_mtp": 77.2, "gmlx_prefill_base": 2679.4, "llama_prefill_base": 1355.5, "range": { "gmlx_decode_base": [ 90.1, 90.9 ], "gmlx_decode_mtp": [ 89.5, 124.4 ], "llama_decode_base": [ 72.7, 78.5 ], "llama_decode_mtp": [ 65.6, 97.0 ], "gmlx_prefill_base": [ 2647.9, 2731.0 ], "llama_prefill_base": [ 1281.5, 1413.2 ] } }, { "depth": 67000, "gmlx_decode_base": 68.6, "gmlx_decode_mtp": 69.7, "llama_decode_base": 50.5, "llama_decode_mtp": 40.2, "gmlx_prefill_base": 1992.6, "llama_prefill_base": 676.3, "range": { "gmlx_decode_base": [ 67.9, 69.3 ], "gmlx_decode_mtp": [ 65.7, 75.5 ], "llama_decode_base": [ 49.4, 51.0 ], "llama_decode_mtp": [ 36.1, 47.8 ], "gmlx_prefill_base": [ 1964.0, 2010.6 ], "llama_prefill_base": [ 622.2, 722.3 ] } }, { "depth": 110000, "gmlx_decode_base": 57.9, "gmlx_decode_mtp": 53.8, "llama_decode_base": 41.3, "llama_decode_mtp": 32.8, "gmlx_prefill_base": 1558.0, "llama_prefill_base": 502.0, "range": { "gmlx_decode_base": [ 57.5, 58.5 ], "gmlx_decode_mtp": [ 49.9, 70.2 ], "llama_decode_base": [ 41.1, 43.0 ], "llama_decode_mtp": [ 30.0, 33.6 ], "gmlx_prefill_base": [ 1541.4, 1564.8 ], "llama_prefill_base": [ 491.3, 533.6 ] } } ] }, "gemma-4-12b-q6k": { "label": "gemma-4-12B-it Q6_K", "comparison": "llama.cpp", "mtp": true, "mtp_kind": "drafter", "gguf": "gemma-4-12b-it-Q6_K.gguf", "source": null, "builds": { "gmlx": "0.1.0", "mlx-kquant": "0.3.5", "llama.cpp": "b9967" }, "measured": "2026-07-18", "rows": [ { "depth": 512, "gmlx_decode_base": 43.9, "gmlx_decode_mtp": 71.9, "llama_decode_base": 39.9, "llama_decode_mtp": 54.0, "gmlx_prefill_base": 854.4, "llama_prefill_base": 726.2, "range": { "gmlx_decode_base": [ 43.3, 44.0 ], "gmlx_decode_mtp": [ 66.3, 87.3 ], "llama_decode_base": [ 38.5, 40.4 ], "llama_decode_mtp": [ 51.4, 55.3 ], "gmlx_prefill_base": [ 808.8, 965.4 ], "llama_prefill_base": [ 589.4, 773.2 ] } }, { "depth": 4300, "gmlx_decode_base": 42.9, "gmlx_decode_mtp": 67.1, "llama_decode_base": 39.8, "llama_decode_mtp": 57.5, "gmlx_prefill_base": 1208.0, "llama_prefill_base": 856.8, "range": { "gmlx_decode_base": [ 42.8, 43.3 ], "gmlx_decode_mtp": [ 62.9, 76.9 ], "llama_decode_base": [ 39.2, 40.8 ], "llama_decode_mtp": [ 52.4, 65.4 ], "gmlx_prefill_base": [ 1185.8, 1228.8 ], "llama_prefill_base": [ 772.5, 899.7 ] } }, { "depth": 17000, "gmlx_decode_base": 40.2, "gmlx_decode_mtp": 67.6, "llama_decode_base": 35.7, "llama_decode_mtp": 52.7, "gmlx_prefill_base": 1219.7, "llama_prefill_base": 771.2, "range": { "gmlx_decode_base": [ 40.0, 40.5 ], "gmlx_decode_mtp": [ 63.8, 77.0 ], "llama_decode_base": [ 35.1, 36.7 ], "llama_decode_mtp": [ 46.9, 56.5 ], "gmlx_prefill_base": [ 1209.2, 1239.5 ], "llama_prefill_base": [ 761.8, 811.0 ] } }, { "depth": 67000, "gmlx_decode_base": 34.4, "gmlx_decode_mtp": 50.1, "llama_decode_base": 26.9, "llama_decode_mtp": 27.5, "gmlx_prefill_base": 961.2, "llama_prefill_base": 431.9, "range": { "gmlx_decode_base": [ 33.9, 34.9 ], "gmlx_decode_mtp": [ 47.1, 66.6 ], "llama_decode_base": [ 26.4, 27.1 ], "llama_decode_mtp": [ 23.7, 33.2 ], "gmlx_prefill_base": [ 944.2, 976.6 ], "llama_prefill_base": [ 411.5, 438.0 ] } }, { "depth": 110000, "gmlx_decode_base": 30.1, "gmlx_decode_mtp": 46.2, "llama_decode_base": 22.3, "llama_decode_mtp": 21.5, "gmlx_prefill_base": 821.0, "llama_prefill_base": 310.1, "range": { "gmlx_decode_base": [ 30.0, 30.7 ], "gmlx_decode_mtp": [ 38.6, 53.1 ], "llama_decode_base": [ 22.2, 22.5 ], "llama_decode_mtp": [ 17.9, 22.8 ], "gmlx_prefill_base": [ 806.1, 827.9 ], "llama_prefill_base": [ 306.7, 326.3 ] } } ] }, "gemma-4-e4b-q6k": { "label": "gemma-4-E4B-it Q6_K", "comparison": "llama.cpp", "mtp": false, "mtp_kind": null, "gguf": "gemma-4-E4B-it-Q6_K.gguf", "source": null, "builds": { "gmlx": "0.1.0", "mlx-kquant": "0.3.5", "llama.cpp": "b9967" }, "measured": "2026-07-18", "rows": [ { "depth": 512, "gmlx_decode_base": 91.9, "gmlx_decode_mtp": null, "llama_decode_base": 92.6, "llama_decode_mtp": null, "gmlx_prefill_base": 2317.7, "llama_prefill_base": 2168.9, "range": { "gmlx_decode_base": [ 91.2, 92.1 ], "gmlx_decode_mtp": null, "llama_decode_base": [ 92.3, 93.5 ], "llama_decode_mtp": null, "gmlx_prefill_base": [ 2223.7, 2456.6 ], "llama_prefill_base": [ 1974.8, 2547.6 ] } }, { "depth": 4300, "gmlx_decode_base": 91.3, "gmlx_decode_mtp": null, "llama_decode_base": 89.9, "llama_decode_mtp": null, "gmlx_prefill_base": 4564.5, "llama_prefill_base": 2984.5, "range": { "gmlx_decode_base": [ 90.2, 91.8 ], "gmlx_decode_mtp": null, "llama_decode_base": [ 85.9, 90.8 ], "llama_decode_mtp": null, "gmlx_prefill_base": [ 4367.5, 4613.2 ], "llama_prefill_base": [ 2769.6, 3247.5 ] } }, { "depth": 17000, "gmlx_decode_base": 84.1, "gmlx_decode_mtp": null, "llama_decode_base": 75.5, "llama_decode_mtp": null, "gmlx_prefill_base": 4810.2, "llama_prefill_base": 2221.4, "range": { "gmlx_decode_base": [ 82.9, 85.0 ], "gmlx_decode_mtp": null, "llama_decode_base": [ 74.3, 81.1 ], "llama_decode_mtp": null, "gmlx_prefill_base": [ 4662.7, 4954.0 ], "llama_prefill_base": [ 2200.4, 2314.9 ] } }, { "depth": 67000, "gmlx_decode_base": 62.8, "gmlx_decode_mtp": null, "llama_decode_base": 57.1, "llama_decode_mtp": null, "gmlx_prefill_base": 4095.1, "llama_prefill_base": 1119.9, "range": { "gmlx_decode_base": [ 61.1, 63.5 ], "gmlx_decode_mtp": null, "llama_decode_base": [ 54.2, 60.9 ], "llama_decode_mtp": null, "gmlx_prefill_base": [ 4031.9, 4239.9 ], "llama_prefill_base": [ 1084.2, 1159.4 ] } }, { "depth": 110000, "gmlx_decode_base": 51.9, "gmlx_decode_mtp": null, "llama_decode_base": 45.7, "llama_decode_mtp": null, "gmlx_prefill_base": 3326.1, "llama_prefill_base": 754.2, "range": { "gmlx_decode_base": [ 51.2, 52.2 ], "gmlx_decode_mtp": null, "llama_decode_base": [ 45.2, 46.0 ], "llama_decode_mtp": null, "gmlx_prefill_base": [ 3266.9, 3420.2 ], "llama_prefill_base": [ 717.8, 783.8 ] } } ] }, "gemma-4-e2b-q6kxl": { "label": "gemma-4-E2B-it UD-Q6_K_XL", "comparison": "llama.cpp", "mtp": false, "mtp_kind": null, "gguf": "gemma-4-E2B-it-UD-Q6_K_XL.gguf", "source": null, "builds": { "gmlx": "0.1.0", "mlx-kquant": "0.3.5", "llama.cpp": "b9967" }, "measured": "2026-07-18", "rows": [ { "depth": 512, "gmlx_decode_base": 155.3, "gmlx_decode_mtp": null, "llama_decode_base": 145.1, "llama_decode_mtp": null, "gmlx_prefill_base": 5217.0, "llama_prefill_base": 3273.8, "range": { "gmlx_decode_base": [ 155.0, 156.0 ], "gmlx_decode_mtp": null, "llama_decode_base": [ 143.7, 146.0 ], "llama_decode_mtp": null, "gmlx_prefill_base": [ 5037.3, 5527.0 ], "llama_prefill_base": [ 3080.6, 3759.2 ] } }, { "depth": 4300, "gmlx_decode_base": 153.9, "gmlx_decode_mtp": null, "llama_decode_base": 140.1, "llama_decode_mtp": null, "gmlx_prefill_base": 16620.1, "llama_prefill_base": 5212.7, "range": { "gmlx_decode_base": [ 153.7, 163.0 ], "gmlx_decode_mtp": null, "llama_decode_base": [ 138.9, 141.5 ], "llama_decode_mtp": null, "gmlx_prefill_base": [ 16156.5, 17589.0 ], "llama_prefill_base": [ 5061.9, 5636.2 ] } }, { "depth": 17000, "gmlx_decode_base": 140.4, "gmlx_decode_mtp": null, "llama_decode_base": 123.4, "llama_decode_mtp": null, "gmlx_prefill_base": 16188.3, "llama_prefill_base": 3469.4, "range": { "gmlx_decode_base": [ 139.6, 141.8 ], "gmlx_decode_mtp": null, "llama_decode_base": [ 120.5, 124.1 ], "llama_decode_mtp": null, "gmlx_prefill_base": [ 15211.5, 19019.1 ], "llama_prefill_base": [ 3412.9, 3619.4 ] } }, { "depth": 67000, "gmlx_decode_base": 107.2, "gmlx_decode_mtp": null, "llama_decode_base": 84.6, "llama_decode_mtp": null, "gmlx_prefill_base": 10403.4, "llama_prefill_base": 1460.4, "range": { "gmlx_decode_base": [ 106.6, 107.7 ], "gmlx_decode_mtp": null, "llama_decode_base": [ 83.4, 85.0 ], "llama_decode_mtp": null, "gmlx_prefill_base": [ 9910.2, 11056.5 ], "llama_prefill_base": [ 1417.8, 1464.6 ] } }, { "depth": 110000, "gmlx_decode_base": 88.6, "gmlx_decode_mtp": null, "llama_decode_base": 65.0, "llama_decode_mtp": null, "gmlx_prefill_base": 7706.8, "llama_prefill_base": 975.4, "range": { "gmlx_decode_base": [ 84.6, 90.2 ], "gmlx_decode_mtp": null, "llama_decode_base": [ 64.3, 65.5 ], "llama_decode_mtp": null, "gmlx_prefill_base": [ 7288.2, 8190.8 ], "llama_prefill_base": [ 966.9, 983.2 ] } } ] }, "gpt-oss-120b-heretic-mxfp4": { "label": "gpt-oss-120b MXFP4", "comparison": "llama.cpp", "mtp": false, "mtp_kind": null, "gguf": "gpt-oss-120b-heretic-v2-MXFP4.gguf", "source": "https://huggingface.co/llmfan46/gpt-oss-120b-heretic-v2-GGUF", "builds": { "gmlx": "0.1.0", "mlx-kquant": "0.3.5", "llama.cpp": "b9967" }, "measured": "2026-07-18", "rows": [ { "depth": 512, "gmlx_decode_base": 89.9, "gmlx_decode_mtp": null, "llama_decode_base": 83.7, "llama_decode_mtp": null, "gmlx_prefill_base": 875.5, "llama_prefill_base": 601.1, "range": { "gmlx_decode_base": [ 87.6, 91.2 ], "gmlx_decode_mtp": null, "llama_decode_base": [ 83.7, 84.1 ], "llama_decode_mtp": null, "gmlx_prefill_base": [ 764.8, 1078.1 ], "llama_prefill_base": [ 581.9, 622.7 ] } }, { "depth": 4300, "gmlx_decode_base": 88.5, "gmlx_decode_mtp": null, "llama_decode_base": 66.3, "llama_decode_mtp": null, "gmlx_prefill_base": 1502.4, "llama_prefill_base": 827.5, "range": { "gmlx_decode_base": [ 86.8, 91.6 ], "gmlx_decode_mtp": null, "llama_decode_base": [ 65.6, 69.6 ], "llama_decode_mtp": null, "gmlx_prefill_base": [ 1423.3, 1558.5 ], "llama_prefill_base": [ 780.3, 870.2 ] } }, { "depth": 17000, "gmlx_decode_base": 77.9, "gmlx_decode_mtp": null, "llama_decode_base": 59.2, "llama_decode_mtp": null, "gmlx_prefill_base": 1658.1, "llama_prefill_base": 709.5, "range": { "gmlx_decode_base": [ 76.7, 78.5 ], "gmlx_decode_mtp": null, "llama_decode_base": [ 58.5, 59.7 ], "llama_decode_mtp": null, "gmlx_prefill_base": [ 1579.1, 1676.7 ], "llama_prefill_base": [ 683.0, 737.3 ] } }, { "depth": 67000, "gmlx_decode_base": 57.0, "gmlx_decode_mtp": null, "llama_decode_base": 40.9, "llama_decode_mtp": null, "gmlx_prefill_base": 1374.5, "llama_prefill_base": 391.8, "range": { "gmlx_decode_base": [ 54.0, 65.9 ], "gmlx_decode_mtp": null, "llama_decode_base": [ 39.6, 41.4 ], "llama_decode_mtp": null, "gmlx_prefill_base": [ 1340.1, 1386.8 ], "llama_prefill_base": [ 386.9, 395.4 ] } }, { "depth": 90000, "gmlx_decode_base": 49.8, "gmlx_decode_mtp": null, "llama_decode_base": 35.9, "llama_decode_mtp": null, "gmlx_prefill_base": 1229.0, "llama_prefill_base": 315.2, "range": { "gmlx_decode_base": [ 48.7, 51.3 ], "gmlx_decode_mtp": null, "llama_decode_base": [ 35.9, 35.9 ], "llama_decode_mtp": null, "gmlx_prefill_base": [ 1219.9, 1241.3 ], "llama_prefill_base": [ 308.4, 321.2 ] } } ] }, "gpt-oss-20b-mxfp4": { "label": "gpt-oss-20b MXFP4", "comparison": "llama.cpp", "mtp": false, "mtp_kind": null, "gguf": "gpt-oss-20b-mxfp4.gguf", "source": null, "builds": { "gmlx": "0.1.0", "mlx-kquant": "0.3.5", "llama.cpp": "b9967" }, "measured": "2026-07-18", "rows": [ { "depth": 512, "gmlx_decode_base": 128.4, "gmlx_decode_mtp": null, "llama_decode_base": 122.0, "llama_decode_mtp": null, "gmlx_prefill_base": 2307.5, "llama_prefill_base": 1235.5, "range": { "gmlx_decode_base": [ 126.5, 130.5 ], "gmlx_decode_mtp": null, "llama_decode_base": [ 116.0, 122.0 ], "llama_decode_mtp": null, "gmlx_prefill_base": [ 2149.1, 2689.5 ], "llama_prefill_base": [ 1055.5, 1666.8 ] } }, { "depth": 4300, "gmlx_decode_base": 133.4, "gmlx_decode_mtp": null, "llama_decode_base": 119.1, "llama_decode_mtp": null, "gmlx_prefill_base": 3441.8, "llama_prefill_base": 1937.4, "range": { "gmlx_decode_base": [ 131.3, 135.1 ], "gmlx_decode_mtp": null, "llama_decode_base": [ 112.9, 121.6 ], "llama_decode_mtp": null, "gmlx_prefill_base": [ 3269.6, 3525.8 ], "llama_prefill_base": [ 1717.7, 1973.9 ] } }, { "depth": 17000, "gmlx_decode_base": 121.7, "gmlx_decode_mtp": null, "llama_decode_base": 107.0, "llama_decode_mtp": null, "gmlx_prefill_base": 3520.6, "llama_prefill_base": 1529.7, "range": { "gmlx_decode_base": [ 119.7, 122.6 ], "gmlx_decode_mtp": null, "llama_decode_base": [ 104.0, 107.8 ], "llama_decode_mtp": null, "gmlx_prefill_base": [ 3354.2, 3607.7 ], "llama_prefill_base": [ 1507.3, 1553.0 ] } }, { "depth": 67000, "gmlx_decode_base": 88.2, "gmlx_decode_mtp": null, "llama_decode_base": 65.8, "llama_decode_mtp": null, "gmlx_prefill_base": 2636.2, "llama_prefill_base": 662.8, "range": { "gmlx_decode_base": [ 84.2, 92.7 ], "gmlx_decode_mtp": null, "llama_decode_base": [ 64.6, 68.6 ], "llama_decode_mtp": null, "gmlx_prefill_base": [ 2453.0, 2655.5 ], "llama_prefill_base": [ 653.4, 703.2 ] } }, { "depth": 90000, "gmlx_decode_base": 77.5, "gmlx_decode_mtp": null, "llama_decode_base": 56.4, "llama_decode_mtp": null, "gmlx_prefill_base": 2285.5, "llama_prefill_base": 520.9, "range": { "gmlx_decode_base": [ 74.8, 82.8 ], "gmlx_decode_mtp": null, "llama_decode_base": [ 55.7, 57.9 ], "llama_decode_mtp": null, "gmlx_prefill_base": [ 2258.9, 2333.2 ], "llama_prefill_base": [ 517.3, 547.0 ] } } ] }, "dolphin3-llama3.1-8b-q6k": { "label": "Dolphin3.0-Llama3.1-8B Q6_K", "comparison": "llama.cpp", "mtp": false, "mtp_kind": null, "gguf": "Dolphin3.0-Llama3.1-8B-abliterated.Q6_K.gguf", "source": "https://huggingface.co/RavichandranJ/Dolphin3-Cyber-8B-GGUF", "builds": { "gmlx": "0.1.0", "mlx-kquant": "0.3.5", "llama.cpp": "b9967" }, "measured": "2026-07-18", "rows": [ { "depth": 512, "gmlx_decode_base": 80.2, "gmlx_decode_mtp": null, "llama_decode_base": 76.7, "llama_decode_mtp": null, "gmlx_prefill_base": 2384.4, "llama_prefill_base": 2290.8, "range": { "gmlx_decode_base": [ 78.9, 82.1 ], "gmlx_decode_mtp": null, "llama_decode_base": [ 75.3, 78.7 ], "llama_decode_mtp": null, "gmlx_prefill_base": [ 2279.7, 2462.6 ], "llama_prefill_base": [ 2206.5, 2481.0 ] } }, { "depth": 4300, "gmlx_decode_base": 71.3, "gmlx_decode_mtp": null, "llama_decode_base": 67.9, "llama_decode_mtp": null, "gmlx_prefill_base": 2120.1, "llama_prefill_base": 1669.2, "range": { "gmlx_decode_base": [ 70.5, 71.8 ], "gmlx_decode_mtp": null, "llama_decode_base": [ 65.3, 70.2 ], "llama_decode_mtp": null, "gmlx_prefill_base": [ 1940.1, 2236.3 ], "llama_prefill_base": [ 1566.4, 1810.8 ] } }, { "depth": 17000, "gmlx_decode_base": 53.4, "gmlx_decode_mtp": null, "llama_decode_base": 50.1, "llama_decode_mtp": null, "gmlx_prefill_base": 1846.2, "llama_prefill_base": 980.1, "range": { "gmlx_decode_base": [ 52.7, 54.6 ], "gmlx_decode_mtp": null, "llama_decode_base": [ 48.9, 50.4 ], "llama_decode_mtp": null, "gmlx_prefill_base": [ 1752.2, 1902.3 ], "llama_prefill_base": [ 943.4, 1006.4 ] } }, { "depth": 67000, "gmlx_decode_base": 28.7, "gmlx_decode_mtp": null, "llama_decode_base": 26.6, "llama_decode_mtp": null, "gmlx_prefill_base": 1065.5, "llama_prefill_base": 326.4, "range": { "gmlx_decode_base": [ 26.7, 29.1 ], "gmlx_decode_mtp": null, "llama_decode_base": [ 23.3, 27.4 ], "llama_decode_mtp": null, "gmlx_prefill_base": [ 1009.8, 1087.3 ], "llama_prefill_base": [ 308.1, 345.4 ] } }, { "depth": 110000, "gmlx_decode_base": 20.7, "gmlx_decode_mtp": null, "llama_decode_base": 17.4, "llama_decode_mtp": null, "gmlx_prefill_base": 765.6, "llama_prefill_base": 205.9, "range": { "gmlx_decode_base": [ 19.4, 21.9 ], "gmlx_decode_mtp": null, "llama_decode_base": [ 16.6, 17.6 ], "llama_decode_mtp": null, "gmlx_prefill_base": [ 742.2, 776.3 ], "llama_prefill_base": [ 200.7, 208.8 ] } } ] }, "deepseek-v4-flash-unsloth-udiq3xxs": { "label": "DeepSeek-V4-Flash UD-IQ3_XXS", "comparison": "llama.cpp", "mtp": false, "mtp_kind": null, "gguf": "DeepSeek-V4-Flash-UD-IQ3_XXS-00001-of-00004.gguf", "source": "https://huggingface.co/unsloth/DeepSeek-V4-Flash-GGUF", "builds": { "gmlx": "0.1.0", "mlx-kquant": "0.3.5", "llama.cpp": "b9967" }, "measured": "2026-07-13", "rows": [ { "depth": 512, "gmlx_decode_base": 34.9, "gmlx_decode_mtp": null, "llama_decode_base": 12.9, "llama_decode_mtp": null, "gmlx_prefill_base": 377.9, "llama_prefill_base": 312.5, "range": { "gmlx_decode_base": [ 34.1, 35.0 ], "gmlx_decode_mtp": null, "llama_decode_base": [ 12.7, 13.1 ], "llama_decode_mtp": null, "gmlx_prefill_base": [ 334.7, 438.4 ], "llama_prefill_base": [ 280.6, 336.8 ] } }, { "depth": 4300, "gmlx_decode_base": 32.5, "gmlx_decode_mtp": null, "llama_decode_base": 12.5, "llama_decode_mtp": null, "gmlx_prefill_base": 572.3, "llama_prefill_base": 356.7, "range": { "gmlx_decode_base": [ 32.3, 32.7 ], "gmlx_decode_mtp": null, "llama_decode_base": [ 12.4, 12.7 ], "llama_decode_mtp": null, "gmlx_prefill_base": [ 567.3, 582.8 ], "llama_prefill_base": [ 342.2, 370.7 ] } }, { "depth": 17000, "gmlx_decode_base": 31.0, "gmlx_decode_mtp": null, "llama_decode_base": 12.4, "llama_decode_mtp": null, "gmlx_prefill_base": 614.1, "llama_prefill_base": 265.7, "range": { "gmlx_decode_base": [ 30.6, 31.8 ], "gmlx_decode_mtp": null, "llama_decode_base": [ 12.0, 12.5 ], "llama_decode_mtp": null, "gmlx_prefill_base": [ 610.0, 619.6 ], "llama_prefill_base": [ 248.9, 279.8 ] } }, { "depth": 67000, "gmlx_decode_base": 30.1, "gmlx_decode_mtp": null, "llama_decode_base": 10.9, "llama_decode_mtp": null, "gmlx_prefill_base": 597.4, "llama_prefill_base": 152.1, "range": { "gmlx_decode_base": [ 29.1, 30.5 ], "gmlx_decode_mtp": null, "llama_decode_base": [ 10.6, 10.9 ], "llama_decode_mtp": null, "gmlx_prefill_base": [ 591.0, 610.5 ], "llama_prefill_base": [ 142.2, 157.1 ] } }, { "depth": 110000, "gmlx_decode_base": 28.3, "gmlx_decode_mtp": null, "llama_decode_base": 9.9, "llama_decode_mtp": null, "gmlx_prefill_base": 543.8, "llama_prefill_base": 116.5, "range": { "gmlx_decode_base": [ 28.0, 28.9 ], "gmlx_decode_mtp": null, "llama_decode_base": [ 9.7, 9.9 ], "llama_decode_mtp": null, "gmlx_prefill_base": [ 492.2, 569.6 ], "llama_prefill_base": [ 112.0, 118.7 ] } }, { "depth": 200000, "gmlx_decode_base": 24.9, "gmlx_decode_mtp": null, "llama_decode_base": null, "llama_decode_mtp": null, "gmlx_prefill_base": 484.3, "llama_prefill_base": null, "range": { "gmlx_decode_base": [ 24.8, 25.1 ], "gmlx_decode_mtp": null, "llama_decode_base": null, "llama_decode_mtp": null, "gmlx_prefill_base": [ 450.0, 493.2 ], "llama_prefill_base": null } }, { "depth": 384000, "gmlx_decode_base": 22.3, "gmlx_decode_mtp": null, "llama_decode_base": null, "llama_decode_mtp": null, "gmlx_prefill_base": 411.1, "llama_prefill_base": null, "range": { "gmlx_decode_base": [ 20.7, 22.9 ], "gmlx_decode_mtp": null, "llama_decode_base": null, "llama_decode_mtp": null, "gmlx_prefill_base": [ 397.2, 431.8 ], "llama_prefill_base": null } } ] }, "qwen38-flash-next-udq3kxl": { "label": "Qwen3.8-Flash-Next UD-Q3_K_XL", "comparison": "llama.cpp", "mtp": true, "mtp_kind": "native", "gguf": "Qwen3.8-Flash-Next-UD-Q3_K_XL-00001-of-00003.gguf", "source": "https://huggingface.co/unsloth/Qwen3.8-Flash-Next-GGUF", "builds": { "gmlx": "0.4.5+#98", "mlx-kquant": "0.4.3", "llama.cpp": "b10636+pr27742" }, "measured": "2026-08-29", "rows": [ { "depth": 512, "gmlx_decode_base": 50.4, "gmlx_decode_mtp": 61.0, "llama_decode_base": 35.4, "llama_decode_mtp": null, "gmlx_prefill_base": 743.0, "llama_prefill_base": 536.1, "range": { "gmlx_decode_base": [ 48.1, 50.8 ], "gmlx_decode_mtp": [ 51.7, 67.5 ], "llama_decode_base": [ 35.2, 35.5 ], "llama_decode_mtp": null, "gmlx_prefill_base": [ 637.0, 842.2 ], "llama_prefill_base": [ 402.5, 567.3 ] } }, { "depth": 4300, "gmlx_decode_base": 47.8, "gmlx_decode_mtp": 54.3, "llama_decode_base": 33.5, "llama_decode_mtp": null, "gmlx_prefill_base": 1338.6, "llama_prefill_base": 688.6, "range": { "gmlx_decode_base": [ 46.2, 48.4 ], "gmlx_decode_mtp": [ 47.5, 57.5 ], "llama_decode_base": [ 33.0, 33.9 ], "llama_decode_mtp": null, "gmlx_prefill_base": [ 1249.5, 1392.2 ], "llama_prefill_base": [ 666.5, 710.7 ] } }, { "depth": 17000, "gmlx_decode_base": 44.9, "gmlx_decode_mtp": 54.1, "llama_decode_base": 29.8, "llama_decode_mtp": null, "gmlx_prefill_base": 1516.1, "llama_prefill_base": 600.3, "range": { "gmlx_decode_base": [ 44.4, 45.9 ], "gmlx_decode_mtp": [ 49.6, 58.3 ], "llama_decode_base": [ 28.1, 30.4 ], "llama_decode_mtp": null, "gmlx_prefill_base": [ 1485.9, 1559.4 ], "llama_prefill_base": [ 573.5, 614.6 ] } }, { "depth": 67000, "gmlx_decode_base": 36.8, "gmlx_decode_mtp": 51.4, "llama_decode_base": 21.1, "llama_decode_mtp": null, "gmlx_prefill_base": 1447.8, "llama_prefill_base": 394.7, "range": { "gmlx_decode_base": [ 36.4, 37.6 ], "gmlx_decode_mtp": [ 47.7, 58.1 ], "llama_decode_base": [ 20.5, 21.2 ], "llama_decode_mtp": null, "gmlx_prefill_base": [ 1408.1, 1489.7 ], "llama_prefill_base": [ 386.0, 412.0 ] } }, { "depth": 110000, "gmlx_decode_base": 31.7, "gmlx_decode_mtp": 49.8, "llama_decode_base": 16.5, "llama_decode_mtp": null, "gmlx_prefill_base": 1379.9, "llama_prefill_base": 288.7, "range": { "gmlx_decode_base": [ 31.0, 32.4 ], "gmlx_decode_mtp": [ 46.0, 55.4 ], "llama_decode_base": [ 16.4, 16.7 ], "llama_decode_mtp": null, "gmlx_prefill_base": [ 1354.6, 1461.2 ], "llama_prefill_base": [ 282.4, 292.4 ] } }, { "depth": 200000, "gmlx_decode_base": 25.3, "gmlx_decode_mtp": 46.8, "llama_decode_base": 11.9, "llama_decode_mtp": null, "gmlx_prefill_base": 1250.0, "llama_prefill_base": 210.8, "range": { "gmlx_decode_base": [ 25.3, 25.6 ], "gmlx_decode_mtp": [ 44.9, 49.7 ], "llama_decode_base": [ 11.8, 12.0 ], "llama_decode_mtp": null, "gmlx_prefill_base": [ 1234.9, 1275.6 ], "llama_prefill_base": [ 207.0, 217.1 ] } } ] }, "deepseek-v4-flash-antirez-iq2xxs": { "label": "DeepSeek-V4-Flash IQ2_XXS", "comparison": "ds4-server", "mtp": false, "mtp_kind": null, "gguf": "DeepSeek-V4-Flash-IQ2XXS-w2Q2K-AProjQ8-SExpQ8-OutQ8-chat-v2-imatrix-0731.gguf", "source": "https://huggingface.co/antirez/deepseek-v4-gguf", "builds": { "gmlx": "0.2.2", "mlx-kquant": "0.3.11", "ds4-server": "b030961" }, "measured": "2026-08-09", "rows": [ { "depth": 512, "gmlx_decode_base": 42.6, "gmlx_decode_mtp": null, "ds4_decode_base": 40.5, "ds4_decode_mtp": null, "gmlx_prefill_base": 541.6, "ds4_prefill_base": 489.7, "range": { "gmlx_decode_base": [ 42.3, 43.1 ], "gmlx_decode_mtp": null, "ds4_decode_base": [ 39.6, 40.7 ], "ds4_decode_mtp": null, "gmlx_prefill_base": [ 541.5, 548.3 ], "ds4_prefill_base": [ 482.0, 492.1 ] } }, { "depth": 4300, "gmlx_decode_base": 38.8, "gmlx_decode_mtp": null, "ds4_decode_base": 33.4, "ds4_decode_mtp": null, "gmlx_prefill_base": 783.5, "ds4_prefill_base": 609.5, "range": { "gmlx_decode_base": [ 37.9, 39.7 ], "gmlx_decode_mtp": null, "ds4_decode_base": [ 33.4, 34.6 ], "ds4_decode_mtp": null, "gmlx_prefill_base": [ 733.9, 803.4 ], "ds4_prefill_base": [ 572.3, 612.9 ] } }, { "depth": 17000, "gmlx_decode_base": 36.9, "gmlx_decode_mtp": null, "ds4_decode_base": 31.4, "ds4_decode_mtp": null, "gmlx_prefill_base": 774.3, "ds4_prefill_base": 567.9, "range": { "gmlx_decode_base": [ 34.8, 37.4 ], "gmlx_decode_mtp": null, "ds4_decode_base": [ 30.8, 31.8 ], "ds4_decode_mtp": null, "gmlx_prefill_base": [ 726.3, 789.7 ], "ds4_prefill_base": [ 524.6, 568.4 ] } }, { "depth": 50000, "gmlx_decode_base": 34.7, "gmlx_decode_mtp": null, "ds4_decode_base": 28.5, "ds4_decode_mtp": null, "gmlx_prefill_base": 765.7, "ds4_prefill_base": 525.7, "range": { "gmlx_decode_base": [ 34.6, 35.2 ], "gmlx_decode_mtp": null, "ds4_decode_base": [ 27.8, 28.5 ], "ds4_decode_mtp": null, "gmlx_prefill_base": [ 756.4, 767.4 ], "ds4_prefill_base": [ 520.6, 525.9 ] } }, { "depth": 67000, "gmlx_decode_base": 33.8, "gmlx_decode_mtp": null, "ds4_decode_base": 27.3, "ds4_decode_mtp": null, "gmlx_prefill_base": 718.2, "ds4_prefill_base": 490.6, "range": { "gmlx_decode_base": [ 32.2, 34.3 ], "gmlx_decode_mtp": null, "ds4_decode_base": [ 27.2, 28.1 ], "ds4_decode_mtp": null, "gmlx_prefill_base": [ 651.3, 725.6 ], "ds4_prefill_base": [ 450.2, 492.4 ] } }, { "depth": 110000, "gmlx_decode_base": 32.2, "gmlx_decode_mtp": null, "ds4_decode_base": 25.1, "ds4_decode_mtp": null, "gmlx_prefill_base": 672.0, "ds4_prefill_base": 432.4, "range": { "gmlx_decode_base": [ 31.8, 32.9 ], "gmlx_decode_mtp": null, "ds4_decode_base": [ 24.1, 25.7 ], "ds4_decode_mtp": null, "gmlx_prefill_base": [ 629.1, 680.0 ], "ds4_prefill_base": [ 411.8, 443.4 ] } }, { "depth": 200000, "gmlx_decode_base": 30.1, "gmlx_decode_mtp": null, "ds4_decode_base": 21.9, "ds4_decode_mtp": null, "gmlx_prefill_base": 602.9, "ds4_prefill_base": 357.1, "range": { "gmlx_decode_base": [ 29.5, 30.2 ], "gmlx_decode_mtp": null, "ds4_decode_base": [ 21.3, 21.9 ], "ds4_decode_mtp": null, "gmlx_prefill_base": [ 583.8, 619.1 ], "ds4_prefill_base": [ 347.8, 365.8 ] } }, { "depth": 300000, "gmlx_decode_base": 27.6, "gmlx_decode_mtp": null, "ds4_decode_base": 18.0, "ds4_decode_mtp": null, "gmlx_prefill_base": 542.4, "ds4_prefill_base": 315.8, "range": { "gmlx_decode_base": [ 27.5, 27.7 ], "gmlx_decode_mtp": null, "ds4_decode_base": [ 18.0, 18.5 ], "ds4_decode_mtp": null, "gmlx_prefill_base": [ 540.8, 542.8 ], "ds4_prefill_base": [ 312.4, 321.2 ] } }, { "depth": 384000, "gmlx_decode_base": 24.9, "gmlx_decode_mtp": null, "ds4_decode_base": 16.2, "ds4_decode_mtp": null, "gmlx_prefill_base": 495.4, "ds4_prefill_base": 275.9, "range": { "gmlx_decode_base": [ 24.8, 25.0 ], "gmlx_decode_mtp": null, "ds4_decode_base": [ 16.2, 16.2 ], "ds4_decode_mtp": null, "gmlx_prefill_base": [ 493.8, 496.9 ], "ds4_prefill_base": [ 275.8, 276.1 ] } }, { "depth": 500000, "gmlx_decode_base": 22.6, "gmlx_decode_mtp": null, "ds4_decode_base": 14.2, "ds4_decode_mtp": null, "gmlx_prefill_base": 440.7, "ds4_prefill_base": 236.9, "range": { "gmlx_decode_base": [ 22.6, 22.7 ], "gmlx_decode_mtp": null, "ds4_decode_base": [ 14.2, 14.2 ], "ds4_decode_mtp": null, "gmlx_prefill_base": [ 438.9, 442.5 ], "ds4_prefill_base": [ 236.5, 237.3 ] } } ] }, "deepseek-v4.1-flash-antirez-q2": { "label": "DeepSeek-V4.1-Flash Q2", "comparison": "ds4-server", "mtp": false, "mtp_kind": null, "gguf": "DeepSeek-V4.1-Flash-Q2.gguf", "source": "https://huggingface.co/antirez/deepseek-v4.1-flash-gguf", "builds": { "gmlx": "0.4.13+ds41", "mlx-kquant": "0.4.11", "ds4-server": "8db1d1d" }, "measured": "2026-09-18", "rows": [ { "depth": 512, "gmlx_decode_base": 20.8, "gmlx_decode_mtp": null, "ds4_decode_base": 15.5, "ds4_decode_mtp": null, "gmlx_prefill_base": 53.3, "ds4_prefill_base": 34.8, "range": { "gmlx_decode_base": [ 18.3, 20.9 ], "gmlx_decode_mtp": null, "ds4_decode_base": [ 15.0, 16.1 ], "ds4_decode_mtp": null, "gmlx_prefill_base": [ 47.1, 84.6 ], "ds4_prefill_base": [ 30.4, 59.2 ] } }, { "depth": 4300, "gmlx_decode_base": 18.1, "gmlx_decode_mtp": null, "ds4_decode_base": 15.1, "ds4_decode_mtp": null, "gmlx_prefill_base": 332.6, "ds4_prefill_base": 113.9, "range": { "gmlx_decode_base": [ 17.6, 19.0 ], "gmlx_decode_mtp": null, "ds4_decode_base": [ 14.2, 16.1 ], "ds4_decode_mtp": null, "gmlx_prefill_base": [ 322.8, 354.5 ], "ds4_prefill_base": [ 76.0, 118.8 ] } }, { "depth": 17000, "gmlx_decode_base": 18.6, "gmlx_decode_mtp": null, "ds4_decode_base": 14.5, "ds4_decode_mtp": null, "gmlx_prefill_base": 611.4, "ds4_prefill_base": 189.9, "range": { "gmlx_decode_base": [ 18.2, 20.6 ], "gmlx_decode_mtp": null, "ds4_decode_base": [ 14.0, 16.1 ], "ds4_decode_mtp": null, "gmlx_prefill_base": [ 485.1, 620.6 ], "ds4_prefill_base": [ 176.5, 335.0 ] } }, { "depth": 50000, "gmlx_decode_base": 18.7, "gmlx_decode_mtp": null, "ds4_decode_base": 14.7, "ds4_decode_mtp": null, "gmlx_prefill_base": 668.8, "ds4_prefill_base": 507.5, "range": { "gmlx_decode_base": [ 17.0, 20.5 ], "gmlx_decode_mtp": null, "ds4_decode_base": [ 13.3, 15.8 ], "ds4_decode_mtp": null, "gmlx_prefill_base": [ 662.9, 685.0 ], "ds4_prefill_base": [ 489.5, 519.0 ] } }, { "depth": 67000, "gmlx_decode_base": 19.2, "gmlx_decode_mtp": null, "ds4_decode_base": 14.9, "ds4_decode_mtp": null, "gmlx_prefill_base": 719.7, "ds4_prefill_base": 511.0, "range": { "gmlx_decode_base": [ 17.1, 21.0 ], "gmlx_decode_mtp": null, "ds4_decode_base": [ 14.5, 15.6 ], "ds4_decode_mtp": null, "gmlx_prefill_base": [ 716.4, 727.8 ], "ds4_prefill_base": [ 483.0, 548.0 ] } }, { "depth": 110000, "gmlx_decode_base": 19.6, "gmlx_decode_mtp": null, "ds4_decode_base": 14.8, "ds4_decode_mtp": null, "gmlx_prefill_base": 798.5, "ds4_prefill_base": 626.8, "range": { "gmlx_decode_base": [ 18.9, 20.5 ], "gmlx_decode_mtp": null, "ds4_decode_base": [ 13.4, 15.1 ], "ds4_decode_mtp": null, "gmlx_prefill_base": [ 759.8, 835.1 ], "ds4_prefill_base": [ 585.8, 632.0 ] } }, { "depth": 200000, "gmlx_decode_base": 18.7, "gmlx_decode_mtp": null, "ds4_decode_base": 13.1, "ds4_decode_mtp": null, "gmlx_prefill_base": 807.9, "ds4_prefill_base": 556.5, "range": { "gmlx_decode_base": [ 18.1, 18.9 ], "gmlx_decode_mtp": null, "ds4_decode_base": [ 12.9, 13.4 ], "ds4_decode_mtp": null, "gmlx_prefill_base": [ 782.1, 831.4 ], "ds4_prefill_base": [ 536.1, 564.9 ] } }, { "depth": 384000, "gmlx_decode_base": 17.4, "gmlx_decode_mtp": null, "ds4_decode_base": 12.2, "ds4_decode_mtp": null, "gmlx_prefill_base": 796.9, "ds4_prefill_base": 612.5, "range": { "gmlx_decode_base": [ 12.9, 18.2 ], "gmlx_decode_mtp": null, "ds4_decode_base": [ 10.6, 12.7 ], "ds4_decode_mtp": null, "gmlx_prefill_base": [ 790.7, 807.7 ], "ds4_prefill_base": [ 592.9, 633.5 ] } } ] } } }