{
  "date": "2026-09-28",
  "hardware": {
    "gpu": "NVIDIA GeForce RTX 5090, 32 GB",
    "cpu": "AMD Ryzen 9 9950X3D",
    "ram_gb": 128,
    "os": "Omarchy / Arch Linux"
  },
  "model": {
    "repo": "orcarouter/Qwen3.8-Flash-Next-Uncensored-GGUF",
    "revision": "0434906af7b5202b676d43f108cf4f73d25691ef",
    "quant": "IQ3_XXS",
    "source_shards": 2,
    "source_size_gb": 85.2
  },
  "controls": {
    "context_tokens": 262144,
    "parallel_requests": 1,
    "prompt_batch": 512,
    "thinking": false,
    "max_tokens": 256,
    "temperature": 0,
    "top_k": 1,
    "top_p": 1,
    "min_p": 0,
    "seed": 42,
    "presence_penalty": 0,
    "frequency_penalty": 0,
    "repetition_penalty": 1
  },
  "engines": {
    "Strata": {
      "commit": "5de09114dac0160a061628100524a5ac75a9314e",
      "compat_bf16": true,
      "kv": "int8",
      "expert_slots": 11930,
      "mtp_window": 4,
      "suffix_lookup": true,
      "runtime_window": 6,
      "version": "0.1.18",
      "pool_workers": 15,
      "host_worker": true,
      "measurement_date": "2026-09-28",
      "kv_resident": 32768,
      "host_kv_gib": 3.09
    },
    "Unsloth": {
      "llama_cpp": "b11160-a3c12db9d",
      "kv": "q8_0",
      "gpu_fit": "auto",
      "threads": 16,
      "speculation": "none,ngram-mod",
      "separate_mtp_draft": false,
      "measurement_date": "2026-09-27"
    }
  },
  "method": "One warmup, then 3 distinct prompts per size; one engine at a time. Decode tok/s = (output_tokens - 1) / (last_text_time - first_text_time). Every run used 262144-token server capacity. The longest prompts contain 261880 tokens, leaving 256 output tokens plus 8 safety tokens. Processes remained warm between requests and speculative lookup state was not explicitly reset; no prompt KV-prefix reuse. The rest of the desktop remained active. These runs measure speed.",
  "runs": [
    {
      "engine": "Strata",
      "prompt_tokens": 127,
      "repeat": 1,
      "output_tokens": 256,
      "ttft_s": 0.41110654699150473,
      "decode_s": 2.79224859806709,
      "decode_tok_s": 91.32424676532081,
      "total_s": 3.2036314799916,
      "finish_reason": "length",
      "reused_tokens": 0,
      "backend_metrics": {
        "projection": null,
        "time": 1790607628.8248014,
        "duration_s": 3.2,
        "finish": "length",
        "prompt_tokens": 127,
        "reused": 0,
        "output_tokens": 256,
        "prompt_ms": 396.8,
        "decode_ms": 2804.4,
        "decode_tok_s": 91.3
      }
    },
    {
      "engine": "Strata",
      "prompt_tokens": 127,
      "repeat": 2,
      "output_tokens": 256,
      "ttft_s": 0.2728294209809974,
      "decode_s": 1.9771044530207291,
      "decode_tok_s": 128.97649368519552,
      "total_s": 2.2500443160533905,
      "finish_reason": "length",
      "reused_tokens": 0,
      "backend_metrics": {
        "projection": null,
        "time": 1790607632.0298212,
        "duration_s": 2.2,
        "finish": "length",
        "prompt_tokens": 127,
        "reused": 0,
        "output_tokens": 256,
        "prompt_ms": 261.3,
        "decode_ms": 1985.9,
        "decode_tok_s": 128.9
      }
    },
    {
      "engine": "Strata",
      "prompt_tokens": 127,
      "repeat": 3,
      "output_tokens": 256,
      "ttft_s": 0.24916199897415936,
      "decode_s": 2.07037151302211,
      "decode_tok_s": 123.16630053887184,
      "total_s": 2.3196177639765665,
      "finish_reason": "length",
      "reused_tokens": 0,
      "backend_metrics": {
        "projection": null,
        "time": 1790607634.2809398,
        "duration_s": 2.3,
        "finish": "length",
        "prompt_tokens": 127,
        "reused": 0,
        "output_tokens": 256,
        "prompt_ms": 238.1,
        "decode_ms": 2079.1,
        "decode_tok_s": 123.1
      }
    },
    {
      "engine": "Strata",
      "prompt_tokens": 2048,
      "repeat": 1,
      "output_tokens": 256,
      "ttft_s": 1.7284144249279052,
      "decode_s": 1.9255124260671437,
      "decode_tok_s": 132.43227960924517,
      "total_s": 3.6540236129658297,
      "finish_reason": "length",
      "reused_tokens": 0,
      "backend_metrics": {
        "projection": null,
        "time": 1790607636.6076758,
        "duration_s": 3.6,
        "finish": "length",
        "prompt_tokens": 2048,
        "reused": 0,
        "output_tokens": 256,
        "prompt_ms": 1710.1,
        "decode_ms": 1935.1,
        "decode_tok_s": 132.3
      }
    },
    {
      "engine": "Strata",
      "prompt_tokens": 2048,
      "repeat": 2,
      "output_tokens": 256,
      "ttft_s": 1.7260007970035076,
      "decode_s": 1.8811293609905988,
      "decode_tok_s": 135.55686561912867,
      "total_s": 3.6072118000593036,
      "finish_reason": "length",
      "reused_tokens": 0,
      "backend_metrics": {
        "projection": null,
        "time": 1790607640.2615645,
        "duration_s": 3.6,
        "finish": "length",
        "prompt_tokens": 2048,
        "reused": 0,
        "output_tokens": 256,
        "prompt_ms": 1710.3,
        "decode_ms": 1890.1,
        "decode_tok_s": 135.4
      }
    },
    {
      "engine": "Strata",
      "prompt_tokens": 2048,
      "repeat": 3,
      "output_tokens": 256,
      "ttft_s": 1.7288892050273716,
      "decode_s": 1.833016101969406,
      "decode_tok_s": 139.11498089188967,
      "total_s": 3.561998978955671,
      "finish_reason": "length",
      "reused_tokens": 0,
      "backend_metrics": {
        "projection": null,
        "time": 1790607643.869496,
        "duration_s": 3.6,
        "finish": "length",
        "prompt_tokens": 2048,
        "reused": 0,
        "output_tokens": 256,
        "prompt_ms": 1713.4,
        "decode_ms": 1842.0,
        "decode_tok_s": 139.0
      }
    },
    {
      "engine": "Strata",
      "prompt_tokens": 8192,
      "repeat": 1,
      "output_tokens": 256,
      "ttft_s": 6.955668293987401,
      "decode_s": 1.8435361629817635,
      "decode_tok_s": 138.32112714705804,
      "total_s": 8.799294829019345,
      "finish_reason": "length",
      "reused_tokens": 0,
      "backend_metrics": {
        "projection": null,
        "time": 1790607647.4464447,
        "duration_s": 8.8,
        "finish": "length",
        "prompt_tokens": 8192,
        "reused": 0,
        "output_tokens": 256,
        "prompt_ms": 6925.5,
        "decode_ms": 1852.8,
        "decode_tok_s": 138.2
      }
    },
    {
      "engine": "Strata",
      "prompt_tokens": 8192,
      "repeat": 2,
      "output_tokens": 256,
      "ttft_s": 6.956696348031983,
      "decode_s": 1.875391838955693,
      "decode_tok_s": 135.971584552696,
      "total_s": 8.83220694900956,
      "finish_reason": "length",
      "reused_tokens": 0,
      "backend_metrics": {
        "projection": null,
        "time": 1790607656.2488124,
        "duration_s": 8.8,
        "finish": "length",
        "prompt_tokens": 8192,
        "reused": 0,
        "output_tokens": 256,
        "prompt_ms": 6925.0,
        "decode_ms": 1884.8,
        "decode_tok_s": 135.8
      }
    },
    {
      "engine": "Strata",
      "prompt_tokens": 8192,
      "repeat": 3,
      "output_tokens": 256,
      "ttft_s": 6.947278963052668,
      "decode_s": 1.7707961929263547,
      "decode_tok_s": 144.0030202338509,
      "total_s": 8.718172608059831,
      "finish_reason": "length",
      "reused_tokens": 0,
      "backend_metrics": {
        "projection": null,
        "time": 1790607665.0834365,
        "duration_s": 8.7,
        "finish": "length",
        "prompt_tokens": 8192,
        "reused": 0,
        "output_tokens": 256,
        "prompt_ms": 6913.5,
        "decode_ms": 1780.7,
        "decode_tok_s": 143.8
      }
    },
    {
      "engine": "Strata",
      "prompt_tokens": 261880,
      "repeat": 1,
      "output_tokens": 256,
      "ttft_s": 242.92730662506074,
      "decode_s": 2.4778385089011863,
      "decode_tok_s": 102.91227579358326,
      "total_s": 245.40523951605428,
      "finish_reason": "length",
      "reused_tokens": 0,
      "backend_metrics": {
        "projection": null,
        "time": 1790607674.3552117,
        "duration_s": 244.8,
        "finish": "length",
        "prompt_tokens": 261880,
        "reused": 0,
        "output_tokens": 256,
        "prompt_ms": 242315.3,
        "decode_ms": 2490.9,
        "decode_tok_s": 102.8
      }
    },
    {
      "engine": "Strata",
      "prompt_tokens": 261880,
      "repeat": 2,
      "output_tokens": 256,
      "ttft_s": 249.4244816970313,
      "decode_s": 2.2143429630668834,
      "decode_tok_s": 115.15831298636002,
      "total_s": 251.63892582198605,
      "finish_reason": "length",
      "reused_tokens": 0,
      "backend_metrics": {
        "projection": null,
        "time": 1790607919.7582638,
        "duration_s": 251.1,
        "finish": "length",
        "prompt_tokens": 261880,
        "reused": 0,
        "output_tokens": 256,
        "prompt_ms": 248819.0,
        "decode_ms": 2227.4,
        "decode_tok_s": 114.9
      }
    },
    {
      "engine": "Strata",
      "prompt_tokens": 261880,
      "repeat": 3,
      "output_tokens": 256,
      "ttft_s": 246.10979042504914,
      "decode_s": 2.2009299689671025,
      "decode_tok_s": 115.8601153128337,
      "total_s": 248.3107972360449,
      "finish_reason": "length",
      "reused_tokens": 0,
      "backend_metrics": {
        "projection": null,
        "time": 1790608171.4023945,
        "duration_s": 247.7,
        "finish": "length",
        "prompt_tokens": 261880,
        "reused": 0,
        "output_tokens": 256,
        "prompt_ms": 245501.4,
        "decode_ms": 2213.6,
        "decode_tok_s": 115.6
      }
    },
    {
      "engine": "Unsloth",
      "prompt_tokens": 127,
      "repeat": 1,
      "output_tokens": 256,
      "ttft_s": 1.0200162520632148,
      "decode_s": 4.888610251946375,
      "decode_tok_s": 52.16206382958696,
      "total_s": 5.909560805070214,
      "finish_reason": "length",
      "reused_tokens": 0,
      "backend_metrics": {
        "cache_n": 0,
        "prompt_n": 127,
        "prompt_ms": 965.728,
        "prompt_per_token_ms": 7.60415748031496,
        "prompt_per_second": 131.50700818449917,
        "predicted_n": 256,
        "predicted_ms": 4888.306,
        "predicted_per_token_ms": 19.16982745098039,
        "predicted_per_second": 52.16531043678526
      }
    },
    {
      "engine": "Unsloth",
      "prompt_tokens": 127,
      "repeat": 2,
      "output_tokens": 256,
      "ttft_s": 0.8814755170606077,
      "decode_s": 6.213637363049202,
      "decode_tok_s": 41.03876443070449,
      "total_s": 7.095767534105107,
      "finish_reason": "length",
      "reused_tokens": 0,
      "backend_metrics": {
        "cache_n": 0,
        "prompt_n": 127,
        "prompt_ms": 827.249,
        "prompt_per_token_ms": 6.513771653543308,
        "prompt_per_second": 153.52088669795913,
        "predicted_n": 256,
        "predicted_ms": 6213.891,
        "predicted_per_token_ms": 24.368199999999998,
        "predicted_per_second": 41.03708932132862,
        "draft_n": 192,
        "draft_n_accepted": 33
      }
    },
    {
      "engine": "Unsloth",
      "prompt_tokens": 127,
      "repeat": 3,
      "output_tokens": 256,
      "ttft_s": 0.9002167250728235,
      "decode_s": 5.623098628944717,
      "decode_tok_s": 45.34866215708823,
      "total_s": 6.5239512691041455,
      "finish_reason": "length",
      "reused_tokens": 0,
      "backend_metrics": {
        "cache_n": 0,
        "prompt_n": 127,
        "prompt_ms": 839.346,
        "prompt_per_token_ms": 6.6090236220472445,
        "prompt_per_second": 151.30828049457554,
        "predicted_n": 256,
        "predicted_ms": 5624.249,
        "predicted_per_token_ms": 22.055878431372548,
        "predicted_per_second": 45.33938664522144,
        "draft_n": 128,
        "draft_n_accepted": 28
      }
    },
    {
      "engine": "Unsloth",
      "prompt_tokens": 2048,
      "repeat": 1,
      "output_tokens": 256,
      "ttft_s": 4.994244789937511,
      "decode_s": 5.518061191076413,
      "decode_tok_s": 46.211883335468585,
      "total_s": 10.512836843030527,
      "finish_reason": "length",
      "reused_tokens": 0,
      "backend_metrics": {
        "cache_n": 0,
        "prompt_n": 2048,
        "prompt_ms": 4935.663,
        "prompt_per_token_ms": 2.40999169921875,
        "prompt_per_second": 414.9391885142888,
        "predicted_n": 256,
        "predicted_ms": 5518.456,
        "predicted_per_token_ms": 21.64100392156863,
        "predicted_per_second": 46.20857718173344,
        "draft_n": 64,
        "draft_n_accepted": 2
      }
    },
    {
      "engine": "Unsloth",
      "prompt_tokens": 2048,
      "repeat": 2,
      "output_tokens": 256,
      "ttft_s": 4.962941317935474,
      "decode_s": 6.298784626065753,
      "decode_tok_s": 40.4840005077732,
      "total_s": 11.2623477979796,
      "finish_reason": "length",
      "reused_tokens": 0,
      "backend_metrics": {
        "cache_n": 0,
        "prompt_n": 2048,
        "prompt_ms": 4896.0,
        "prompt_per_token_ms": 2.390625,
        "prompt_per_second": 418.30065359477123,
        "predicted_n": 256,
        "predicted_ms": 6299.101,
        "predicted_per_token_ms": 24.7023568627451,
        "predicted_per_second": 40.481967188651204,
        "draft_n": 192,
        "draft_n_accepted": 22
      }
    },
    {
      "engine": "Unsloth",
      "prompt_tokens": 2048,
      "repeat": 3,
      "output_tokens": 256,
      "ttft_s": 4.981224590912461,
      "decode_s": 5.671348358038813,
      "decode_tok_s": 44.962852553141445,
      "total_s": 10.65320850291755,
      "finish_reason": "length",
      "reused_tokens": 0,
      "backend_metrics": {
        "cache_n": 0,
        "prompt_n": 2048,
        "prompt_ms": 4912.442,
        "prompt_per_token_ms": 2.3986533203125,
        "prompt_per_second": 416.900596485414,
        "predicted_n": 256,
        "predicted_ms": 5671.805,
        "predicted_per_token_ms": 22.24237254901961,
        "predicted_per_second": 44.959232554715825,
        "draft_n": 128,
        "draft_n_accepted": 15
      }
    },
    {
      "engine": "Unsloth",
      "prompt_tokens": 8192,
      "repeat": 1,
      "output_tokens": 256,
      "ttft_s": 19.322279473068193,
      "decode_s": 5.839794173953123,
      "decode_tok_s": 43.66592253154416,
      "total_s": 25.162929136073217,
      "finish_reason": "length",
      "reused_tokens": 0,
      "backend_metrics": {
        "cache_n": 0,
        "prompt_n": 8192,
        "prompt_ms": 19250.134,
        "prompt_per_token_ms": 2.349869873046875,
        "prompt_per_second": 425.55547925017044,
        "predicted_n": 256,
        "predicted_ms": 5839.966,
        "predicted_per_token_ms": 22.901827450980395,
        "predicted_per_second": 43.664637773576075,
        "draft_n": 320,
        "draft_n_accepted": 117
      }
    },
    {
      "engine": "Unsloth",
      "prompt_tokens": 8192,
      "repeat": 2,
      "output_tokens": 256,
      "ttft_s": 19.229636878008023,
      "decode_s": 4.1666103129973635,
      "decode_tok_s": 61.200827733889724,
      "total_s": 23.39681788405869,
      "finish_reason": "length",
      "reused_tokens": 0,
      "backend_metrics": {
        "cache_n": 0,
        "prompt_n": 8192,
        "prompt_ms": 19148.194,
        "prompt_per_token_ms": 2.337426025390625,
        "prompt_per_second": 427.8210258366925,
        "predicted_n": 256,
        "predicted_ms": 4167.257,
        "predicted_per_token_ms": 16.34218431372549,
        "predicted_per_second": 61.19133041230719,
        "draft_n": 256,
        "draft_n_accepted": 165
      }
    },
    {
      "engine": "Unsloth",
      "prompt_tokens": 8192,
      "repeat": 3,
      "output_tokens": 256,
      "ttft_s": 19.269358508987352,
      "decode_s": 2.726838508970104,
      "decode_tok_s": 93.51488882130779,
      "total_s": 21.99688732402865,
      "finish_reason": "length",
      "reused_tokens": 0,
      "backend_metrics": {
        "cache_n": 0,
        "prompt_n": 8192,
        "prompt_ms": 19172.438,
        "prompt_per_token_ms": 2.340385498046875,
        "prompt_per_second": 427.2800360600984,
        "predicted_n": 256,
        "predicted_ms": 2727.122,
        "predicted_per_token_ms": 10.694596078431372,
        "predicted_per_second": 93.50516771893594,
        "draft_n": 192,
        "draft_n_accepted": 192
      }
    },
    {
      "engine": "Unsloth",
      "prompt_tokens": 261880,
      "repeat": 1,
      "output_tokens": 256,
      "ttft_s": 630.8476773110451,
      "decode_s": 10.34023711795453,
      "decode_tok_s": 24.66094317675021,
      "total_s": 641.1898207879858,
      "finish_reason": "length",
      "reused_tokens": 0,
      "backend_metrics": {
        "cache_n": 0,
        "prompt_n": 261880,
        "prompt_ms": 630503.717,
        "prompt_per_token_ms": 2.407605456697724,
        "prompt_per_second": 415.35044590387406,
        "predicted_n": 256,
        "predicted_ms": 10340.205,
        "predicted_per_token_ms": 40.54982352941177,
        "predicted_per_second": 24.661019776687212,
        "draft_n": 64,
        "draft_n_accepted": 25
      }
    },
    {
      "engine": "Unsloth",
      "prompt_tokens": 261880,
      "repeat": 2,
      "output_tokens": 256,
      "ttft_s": 632.0597221710486,
      "decode_s": 11.24783979903441,
      "decode_tok_s": 22.671019907475117,
      "total_s": 643.3095331210643,
      "finish_reason": "length",
      "reused_tokens": 0,
      "backend_metrics": {
        "cache_n": 0,
        "prompt_n": 261880,
        "prompt_ms": 631034.894,
        "prompt_per_token_ms": 2.409633778829998,
        "prompt_per_second": 415.0008224426334,
        "predicted_n": 256,
        "predicted_ms": 11247.733,
        "predicted_per_token_ms": 44.108756862745096,
        "predicted_per_second": 22.67123517245653,
        "draft_n": 117,
        "draft_n_accepted": 14
      }
    },
    {
      "engine": "Unsloth",
      "prompt_tokens": 261880,
      "repeat": 3,
      "output_tokens": 256,
      "ttft_s": 646.5290633520344,
      "decode_s": 11.508479923009872,
      "decode_tok_s": 22.15757438913866,
      "total_s": 658.0394942710409,
      "finish_reason": "length",
      "reused_tokens": 0,
      "backend_metrics": {
        "cache_n": 0,
        "prompt_n": 261880,
        "prompt_ms": 645746.028,
        "prompt_per_token_ms": 2.4658088742935695,
        "prompt_per_second": 405.5464356646418,
        "predicted_n": 256,
        "predicted_ms": 11508.286,
        "predicted_per_token_ms": 45.13053333333333,
        "predicted_per_second": 22.15794776042236,
        "draft_n": 182,
        "draft_n_accepted": 46
      }
    }
  ]
}
