{
  "runs": [
    {
      "meta": {
        "measured": "2026-10-05",
        "ollama": "0.35.0",
        "note": "One machine, one model. Median of 3 warm runs per point after one discarded load run, 96 generated tokens, temperature 0, Windows desktop active."
      },
      "machine": {
        "gpu": "NVIDIA GeForce RTX 3050",
        "gpuMemoryMiB": 6144,
        "gpuDesktopUsedMiB": 1155,
        "cpu": "Intel(R) Xeon(R) CPU E3-1245 v5 @ 3.50GHz",
        "cpuThreads": 8,
        "systemRamGiB": 32,
        "os": "Windows_NT 10.0.19045"
      },
      "model": {
        "tag": "qwen2.5:7b-instruct-q4_K_M",
        "family": "qwen2",
        "parameterSize": "7.6B",
        "quantization": "Q4_K_M",
        "architecture": "qwen2",
        "blocks": 28,
        "kvHeads": 4,
        "heads": 28,
        "contextMax": 32768,
        "embedding": 3584,
        "parameterCount": 7615616512
      },
      "contextSweep": [
        {
          "numCtx": 2048,
          "genTps": 18.951297340573763,
          "promptTps": 928.2276341051081,
          "genTpsRuns": [
            19.26,
            18.79,
            18.95
          ],
          "promptTokens": 759,
          "loadSeconds": 0.0053473,
          "totalGiB": 4.659531247802079,
          "vramGiB": 3.929492186754942,
          "gpuShare": 0.8433234971025252,
          "contextLength": 2048,
          "driverUsedMiB": 5265,
          "driverDeltaGiB": 4.013671875,
          "gpu": {
            "name": "NVIDIA GeForce RTX 3050",
            "totalMiB": 6144,
            "usedMiB": 5265,
            "utilization": 57,
            "smClock": 1792,
            "memClock": 6801,
            "powerW": 60.03,
            "tempC": 60
          },
          "layers": null
        },
        {
          "numCtx": 4096,
          "genTps": 19.431448557151235,
          "promptTps": 914.3814149571143,
          "genTpsRuns": [
            19.1,
            19.51,
            19.43
          ],
          "promptTokens": 761,
          "loadSeconds": 0.0036912,
          "totalGiB": 4.771943356841803,
          "vramGiB": 4.028232420794666,
          "gpuShare": 0.8441492531589175,
          "contextLength": 4096,
          "driverUsedMiB": 5187,
          "driverDeltaGiB": 3.9375,
          "gpu": {
            "name": "NVIDIA GeForce RTX 3050",
            "totalMiB": 6144,
            "usedMiB": 5187,
            "utilization": 58,
            "smClock": 1905,
            "memClock": 6801,
            "powerW": 60.22,
            "tempC": 59
          },
          "layers": null
        },
        {
          "numCtx": 8192,
          "genTps": 16.021469820147182,
          "promptTps": 829.2037022470437,
          "genTpsRuns": [
            16.2,
            15.97,
            16.02
          ],
          "promptTokens": 759,
          "loadSeconds": 0.0036464,
          "totalGiB": 5.009130856953561,
          "vramGiB": 3.962558592669666,
          "gpuShare": 0.7910670944379368,
          "contextLength": 8192,
          "driverUsedMiB": 5113,
          "driverDeltaGiB": 3.865234375,
          "gpu": {
            "name": "NVIDIA GeForce RTX 3050",
            "totalMiB": 6144,
            "usedMiB": 5113,
            "utilization": 40,
            "smClock": 1822,
            "memClock": 6801,
            "powerW": 53.31,
            "tempC": 58
          },
          "layers": null
        },
        {
          "numCtx": 16384,
          "genTps": 13.717328583989405,
          "promptTps": 750.3391809175238,
          "genTpsRuns": [
            13.7,
            13.75,
            13.72
          ],
          "promptTokens": 758,
          "loadSeconds": 0.0032212,
          "totalGiB": 5.4139941381290555,
          "vramGiB": 3.9579882798716426,
          "gpuShare": 0.7310662292736481,
          "contextLength": 16384,
          "driverUsedMiB": 5115,
          "driverDeltaGiB": 3.8671875,
          "gpu": {
            "name": "NVIDIA GeForce RTX 3050",
            "totalMiB": 6144,
            "usedMiB": 5115,
            "utilization": 49,
            "smClock": 1837,
            "memClock": 6801,
            "powerW": 48.08,
            "tempC": 55
          },
          "layers": null
        },
        {
          "numCtx": 32768,
          "genTps": 10.069816846539458,
          "promptTps": 604.3890766641231,
          "genTpsRuns": [
            10.07,
            9.99,
            10.19
          ],
          "promptTokens": 758,
          "loadSeconds": 0.0048988,
          "totalGiB": 6.351494139060378,
          "vramGiB": 3.905849608592689,
          "gpuShare": 0.6149497304220939,
          "contextLength": 32768,
          "driverUsedMiB": 5053,
          "driverDeltaGiB": 3.806640625,
          "gpu": {
            "name": "NVIDIA GeForce RTX 3050",
            "totalMiB": 6144,
            "usedMiB": 5053,
            "utilization": 35,
            "smClock": 915,
            "memClock": 5001,
            "powerW": 26.35,
            "tempC": 49
          },
          "layers": null
        }
      ],
      "offloadSweep": [
        {
          "numGpu": 0,
          "numLayers": 29,
          "genTps": 4.7647330407592765,
          "promptTps": 19.44238331854465,
          "genTpsRuns": [
            4.34,
            4.85,
            4.76
          ],
          "totalGiB": 4.7148828115314245,
          "vramGiB": 0,
          "gpuShare": 0,
          "contextLength": 4096,
          "driverUsedMiB": 972,
          "driverDeltaGiB": -0.1787109375,
          "gpu": {
            "name": "NVIDIA GeForce RTX 3050",
            "totalMiB": 6144,
            "usedMiB": 972,
            "utilization": 25,
            "smClock": 210,
            "memClock": 405,
            "powerW": 9.19,
            "tempC": 37
          }
        },
        {
          "numGpu": 6,
          "numLayers": 29,
          "genTps": 5.3560387229307125,
          "promptTps": 343.779207803788,
          "genTpsRuns": [
            5.36,
            5.3,
            5.44
          ],
          "totalGiB": 4.77195312269032,
          "vramGiB": 1.328945311717689,
          "gpuShare": 0.2784908563746451,
          "contextLength": 4096,
          "driverUsedMiB": 2455,
          "driverDeltaGiB": 1.26953125,
          "gpu": {
            "name": "NVIDIA GeForce RTX 3050",
            "totalMiB": 6144,
            "usedMiB": 2455,
            "utilization": 66,
            "smClock": 225,
            "memClock": 5001,
            "powerW": 20.46,
            "tempC": 43
          }
        },
        {
          "numGpu": 12,
          "numLayers": 29,
          "genTps": 7.231402019481633,
          "promptTps": 419.91416691981806,
          "genTpsRuns": [
            7.21,
            7.23,
            7.26
          ],
          "totalGiB": 4.77195312269032,
          "vramGiB": 2.14208007697016,
          "gpuShare": 0.44888958920923,
          "contextLength": 4096,
          "driverUsedMiB": 3316,
          "driverDeltaGiB": 2.1103515625,
          "gpu": {
            "name": "NVIDIA GeForce RTX 3050",
            "totalMiB": 6144,
            "usedMiB": 3316,
            "utilization": 20,
            "smClock": 472,
            "memClock": 5001,
            "powerW": 21.83,
            "tempC": 48
          }
        },
        {
          "numGpu": 17,
          "numLayers": 29,
          "genTps": 9.620768264992108,
          "promptTps": 517.6632357996482,
          "genTpsRuns": [
            9.38,
            9.62,
            9.63
          ],
          "totalGiB": 4.77195312269032,
          "vramGiB": 2.824833983555436,
          "gpuShare": 0.5919659960873334,
          "contextLength": 4096,
          "driverUsedMiB": 4004,
          "driverDeltaGiB": 2.7822265625,
          "gpu": {
            "name": "NVIDIA GeForce RTX 3050",
            "totalMiB": 6144,
            "usedMiB": 4004,
            "utilization": 32,
            "smClock": 750,
            "memClock": 5001,
            "powerW": 24.72,
            "tempC": 52
          }
        },
        {
          "numGpu": 23,
          "numLayers": 29,
          "genTps": 14.723180778129029,
          "promptTps": 739.4001054308419,
          "genTpsRuns": [
            14.83,
            14.72,
            14.72
          ],
          "totalGiB": 4.77195312269032,
          "vramGiB": 3.621669921092689,
          "gpuShare": 0.7589491824368285,
          "contextLength": 4096,
          "driverUsedMiB": 4809,
          "driverDeltaGiB": 3.568359375,
          "gpu": {
            "name": "NVIDIA GeForce RTX 3050",
            "totalMiB": 6144,
            "usedMiB": 4809,
            "utilization": 54,
            "smClock": 1762,
            "memClock": 6801,
            "powerW": 50.61,
            "tempC": 59
          }
        },
        {
          "numGpu": 29,
          "numLayers": 29,
          "genTps": 29.53830289918443,
          "promptTps": 1127.1058387604653,
          "genTpsRuns": [
            29.49,
            29.64,
            29.54
          ],
          "totalGiB": 4.421972654759884,
          "vramGiB": 4.421972654759884,
          "gpuShare": 1,
          "contextLength": 4096,
          "driverUsedMiB": 5496,
          "driverDeltaGiB": 4.2392578125,
          "gpu": {
            "name": "NVIDIA GeForce RTX 3050",
            "totalMiB": 6144,
            "usedMiB": 5496,
            "utilization": 96,
            "smClock": 1522,
            "memClock": 6801,
            "powerW": 69.19,
            "tempC": 63
          }
        }
      ]
    }
  ]
}