{
  "completed_utc": "2026-10-11T12:09:32.722315+00:00",
  "checks": [
    {
      "run_id": "dsv41_flash_fp8__trial_tp8_dspark__32k__vllm__dsparkfix__1011-112022",
      "state": "complete",
      "structured": {
        "json": "24/24",
        "tool": "8/8"
      },
      "retrieval": {
        "passed": 9,
        "total": 9
      }
    },
    {
      "run_id": "dsv41_flash__trial_tp4_a__32k__vllm__native__1011-113945",
      "state": "complete",
      "structured": {
        "json": "24/24",
        "tool": "8/8"
      },
      "retrieval": {
        "passed": 9,
        "total": 9
      }
    },
    {
      "run_id": "dsv41_flash__trial_tp4_a_dspark__32k__vllm__native__1011-114616",
      "state": "complete",
      "structured": {
        "json": "24/24",
        "tool": "8/8"
      },
      "retrieval": {
        "passed": 9,
        "total": 9
      }
    },
    {
      "run_id": "dsv41_flash__trial_tp4_2rep__32k__vllm__native__1011-115241",
      "state": "complete",
      "structured": {
        "json": "24/24",
        "tool": "8/8"
      },
      "retrieval": {
        "passed": 9,
        "total": 9
      }
    }
  ],
  "results": [
    {
      "run_id": "dsv41_flash_fp8__trial_tp8_dspark__32k__vllm__dsparkfix__1011-112022",
      "model": "dsv41_flash_fp8",
      "hf_repo": "timur136/DeepSeek-V4.1-Flash-FP8",
      "model_revision": "5eff209b3d43e46d9c1534f2e9fa0744f2e81bc4",
      "engine_version": "0.30.1rc1.dev616+g1a001d584",
      "image_digest": "h200/dsv41-dense-scale-rescue@sha256:d7bda81f69f56a15859bbc4ea3511a114383cb58c99b95e3257ec766418f7352",
      "layout": "trial_tp8_dspark",
      "gpus": "8",
      "max_model_len": "32768",
      "max_num_seqs": "64",
      "max_num_batched_tokens": "8192",
      "prefix_caching": "False",
      "spec_method": "dspark",
      "spec_num_tokens": "5",
      "workload": "spec_chat",
      "dataset": "h200_trial_chat_c1",
      "concurrency": "1",
      "num_prompts": "8",
      "completed": "8",
      "failed": "0",
      "out_tok_s": "64.14684672159348",
      "ttft_p50_ms": "209.18477450322825",
      "tpot_p50_ms": "14.491112086255395",
      "spec_accept_len": "3.553448275862069",
      "spec_accept_rate": "0.5106896551724138",
      "prefix_hit_rate": "0.0"
    },
    {
      "run_id": "dsv41_flash_fp8__trial_tp8_dspark__32k__vllm__dsparkfix__1011-112022",
      "model": "dsv41_flash_fp8",
      "hf_repo": "timur136/DeepSeek-V4.1-Flash-FP8",
      "model_revision": "5eff209b3d43e46d9c1534f2e9fa0744f2e81bc4",
      "engine_version": "0.30.1rc1.dev616+g1a001d584",
      "image_digest": "h200/dsv41-dense-scale-rescue@sha256:d7bda81f69f56a15859bbc4ea3511a114383cb58c99b95e3257ec766418f7352",
      "layout": "trial_tp8_dspark",
      "gpus": "8",
      "max_model_len": "32768",
      "max_num_seqs": "64",
      "max_num_batched_tokens": "8192",
      "prefix_caching": "False",
      "spec_method": "dspark",
      "spec_num_tokens": "5",
      "workload": "spec_chat",
      "dataset": "h200_trial_chat_c16",
      "concurrency": "16",
      "num_prompts": "32",
      "completed": "32",
      "failed": "0",
      "out_tok_s": "367.9878829585319",
      "ttft_p50_ms": "401.2245295016328",
      "tpot_p50_ms": "21.438135802967217",
      "spec_accept_len": "3.2721571906354514",
      "spec_accept_rate": "0.4544314381270903",
      "prefix_hit_rate": "0.03722661703117729"
    },
    {
      "run_id": "dsv41_flash__trial_tp4_a__32k__vllm__native__1011-113945",
      "model": "dsv41_flash",
      "hf_repo": "deepseek-ai/DeepSeek-V4.1-Flash",
      "model_revision": "2cba9e42aa026125f3ed06c6d98c1db82f7ca027",
      "engine_version": "0.30.1rc1.dev616+g1a001d584",
      "image_digest": "h200/dsv41-dense-scale-rescue@sha256:d7bda81f69f56a15859bbc4ea3511a114383cb58c99b95e3257ec766418f7352",
      "layout": "trial_tp4_a",
      "gpus": "4",
      "max_model_len": "32768",
      "max_num_seqs": "32",
      "max_num_batched_tokens": "8192",
      "prefix_caching": "False",
      "spec_method": "",
      "spec_num_tokens": "",
      "workload": "spec_chat",
      "dataset": "h200_trial_chat_c1",
      "concurrency": "1",
      "num_prompts": "8",
      "completed": "8",
      "failed": "0",
      "out_tok_s": "115.57084006751226",
      "ttft_p50_ms": "114.85109949717298",
      "tpot_p50_ms": "7.954402197969567",
      "spec_accept_len": "",
      "spec_accept_rate": "",
      "prefix_hit_rate": ""
    },
    {
      "run_id": "dsv41_flash__trial_tp4_a__32k__vllm__native__1011-113945",
      "model": "dsv41_flash",
      "hf_repo": "deepseek-ai/DeepSeek-V4.1-Flash",
      "model_revision": "2cba9e42aa026125f3ed06c6d98c1db82f7ca027",
      "engine_version": "0.30.1rc1.dev616+g1a001d584",
      "image_digest": "h200/dsv41-dense-scale-rescue@sha256:d7bda81f69f56a15859bbc4ea3511a114383cb58c99b95e3257ec766418f7352",
      "layout": "trial_tp4_a",
      "gpus": "4",
      "max_model_len": "32768",
      "max_num_seqs": "32",
      "max_num_batched_tokens": "8192",
      "prefix_caching": "False",
      "spec_method": "",
      "spec_num_tokens": "",
      "workload": "spec_chat",
      "dataset": "h200_trial_chat_c16",
      "concurrency": "16",
      "num_prompts": "32",
      "completed": "32",
      "failed": "0",
      "out_tok_s": "901.1042135641459",
      "ttft_p50_ms": "417.7236669929698",
      "tpot_p50_ms": "15.368603205869851",
      "spec_accept_len": "",
      "spec_accept_rate": "",
      "prefix_hit_rate": ""
    },
    {
      "run_id": "dsv41_flash__trial_tp4_a__32k__vllm__native__1011-113945",
      "model": "dsv41_flash",
      "hf_repo": "deepseek-ai/DeepSeek-V4.1-Flash",
      "model_revision": "2cba9e42aa026125f3ed06c6d98c1db82f7ca027",
      "engine_version": "0.30.1rc1.dev616+g1a001d584",
      "image_digest": "h200/dsv41-dense-scale-rescue@sha256:d7bda81f69f56a15859bbc4ea3511a114383cb58c99b95e3257ec766418f7352",
      "layout": "trial_tp4_a",
      "gpus": "4",
      "max_model_len": "32768",
      "max_num_seqs": "32",
      "max_num_batched_tokens": "8192",
      "prefix_caching": "False",
      "spec_method": "",
      "spec_num_tokens": "",
      "workload": "random",
      "dataset": "",
      "concurrency": "16",
      "num_prompts": "64",
      "completed": "64",
      "failed": "0",
      "out_tok_s": "1012.6950893945991",
      "ttft_p50_ms": "466.2489175098017",
      "tpot_p50_ms": "13.960386584307452",
      "spec_accept_len": "",
      "spec_accept_rate": "",
      "prefix_hit_rate": ""
    },
    {
      "run_id": "dsv41_flash__trial_tp4_a__32k__vllm__native__1011-113945",
      "model": "dsv41_flash",
      "hf_repo": "deepseek-ai/DeepSeek-V4.1-Flash",
      "model_revision": "2cba9e42aa026125f3ed06c6d98c1db82f7ca027",
      "engine_version": "0.30.1rc1.dev616+g1a001d584",
      "image_digest": "h200/dsv41-dense-scale-rescue@sha256:d7bda81f69f56a15859bbc4ea3511a114383cb58c99b95e3257ec766418f7352",
      "layout": "trial_tp4_a",
      "gpus": "4",
      "max_model_len": "32768",
      "max_num_seqs": "32",
      "max_num_batched_tokens": "8192",
      "prefix_caching": "False",
      "spec_method": "",
      "spec_num_tokens": "",
      "workload": "random",
      "dataset": "",
      "concurrency": "64",
      "num_prompts": "128",
      "completed": "128",
      "failed": "0",
      "out_tok_s": "1427.8969473393877",
      "ttft_p50_ms": "6317.589833997772",
      "tpot_p50_ms": "19.958024752992845",
      "spec_accept_len": "",
      "spec_accept_rate": "",
      "prefix_hit_rate": ""
    },
    {
      "run_id": "dsv41_flash__trial_tp4_a_dspark__32k__vllm__native__1011-114616",
      "model": "dsv41_flash",
      "hf_repo": "deepseek-ai/DeepSeek-V4.1-Flash",
      "model_revision": "2cba9e42aa026125f3ed06c6d98c1db82f7ca027",
      "engine_version": "0.30.1rc1.dev616+g1a001d584",
      "image_digest": "h200/dsv41-dense-scale-rescue@sha256:d7bda81f69f56a15859bbc4ea3511a114383cb58c99b95e3257ec766418f7352",
      "layout": "trial_tp4_a_dspark",
      "gpus": "4",
      "max_model_len": "32768",
      "max_num_seqs": "32",
      "max_num_batched_tokens": "8192",
      "prefix_caching": "False",
      "spec_method": "dspark",
      "spec_num_tokens": "5",
      "workload": "spec_chat",
      "dataset": "h200_trial_chat_c1",
      "concurrency": "1",
      "num_prompts": "8",
      "completed": "8",
      "failed": "0",
      "out_tok_s": "268.8252817341493",
      "ttft_p50_ms": "118.43733298883308",
      "tpot_p50_ms": "2.973175568645363",
      "spec_accept_len": "3.5403087478559176",
      "spec_accept_rate": "0.5080617495711836",
      "prefix_hit_rate": ""
    },
    {
      "run_id": "dsv41_flash__trial_tp4_a_dspark__32k__vllm__native__1011-114616",
      "model": "dsv41_flash",
      "hf_repo": "deepseek-ai/DeepSeek-V4.1-Flash",
      "model_revision": "2cba9e42aa026125f3ed06c6d98c1db82f7ca027",
      "engine_version": "0.30.1rc1.dev616+g1a001d584",
      "image_digest": "h200/dsv41-dense-scale-rescue@sha256:d7bda81f69f56a15859bbc4ea3511a114383cb58c99b95e3257ec766418f7352",
      "layout": "trial_tp4_a_dspark",
      "gpus": "4",
      "max_model_len": "32768",
      "max_num_seqs": "32",
      "max_num_batched_tokens": "8192",
      "prefix_caching": "False",
      "spec_method": "dspark",
      "spec_num_tokens": "5",
      "workload": "spec_chat",
      "dataset": "h200_trial_chat_c16",
      "concurrency": "16",
      "num_prompts": "32",
      "completed": "32",
      "failed": "0",
      "out_tok_s": "1030.5306605008625",
      "ttft_p50_ms": "272.33933049137704",
      "tpot_p50_ms": "11.6476056430781",
      "spec_accept_len": "3.243467440895894",
      "spec_accept_rate": "0.44869348817917876",
      "prefix_hit_rate": ""
    },
    {
      "run_id": "dsv41_flash__trial_tp4_2rep__32k__vllm__native__1011-115241",
      "model": "dsv41_flash",
      "hf_repo": "deepseek-ai/DeepSeek-V4.1-Flash",
      "model_revision": "2cba9e42aa026125f3ed06c6d98c1db82f7ca027",
      "engine_version": "0.30.1rc1.dev616+g1a001d584",
      "image_digest": "h200/dsv41-dense-scale-rescue@sha256:d7bda81f69f56a15859bbc4ea3511a114383cb58c99b95e3257ec766418f7352",
      "layout": "trial_tp4_2rep",
      "gpus": "8",
      "max_model_len": "32768",
      "max_num_seqs": "32",
      "max_num_batched_tokens": "8192",
      "prefix_caching": "False",
      "spec_method": "",
      "spec_num_tokens": "",
      "workload": "spec_chat",
      "dataset": "h200_trial_chat_c1",
      "concurrency": "1",
      "num_prompts": "8",
      "completed": "8",
      "failed": "0",
      "out_tok_s": "111.35116850525912",
      "ttft_p50_ms": "128.41936349286698",
      "tpot_p50_ms": "7.959138484347576",
      "spec_accept_len": "",
      "spec_accept_rate": "",
      "prefix_hit_rate": ""
    },
    {
      "run_id": "dsv41_flash__trial_tp4_2rep__32k__vllm__native__1011-115241",
      "model": "dsv41_flash",
      "hf_repo": "deepseek-ai/DeepSeek-V4.1-Flash",
      "model_revision": "2cba9e42aa026125f3ed06c6d98c1db82f7ca027",
      "engine_version": "0.30.1rc1.dev616+g1a001d584",
      "image_digest": "h200/dsv41-dense-scale-rescue@sha256:d7bda81f69f56a15859bbc4ea3511a114383cb58c99b95e3257ec766418f7352",
      "layout": "trial_tp4_2rep",
      "gpus": "8",
      "max_model_len": "32768",
      "max_num_seqs": "32",
      "max_num_batched_tokens": "8192",
      "prefix_caching": "False",
      "spec_method": "",
      "spec_num_tokens": "",
      "workload": "spec_chat",
      "dataset": "h200_trial_chat_c16",
      "concurrency": "16",
      "num_prompts": "32",
      "completed": "32",
      "failed": "0",
      "out_tok_s": "1160.4660172880701",
      "ttft_p50_ms": "289.4229554949561",
      "tpot_p50_ms": "11.661997215660271",
      "spec_accept_len": "",
      "spec_accept_rate": "",
      "prefix_hit_rate": ""
    },
    {
      "run_id": "dsv41_flash__trial_tp4_2rep__32k__vllm__native__1011-115241",
      "model": "dsv41_flash",
      "hf_repo": "deepseek-ai/DeepSeek-V4.1-Flash",
      "model_revision": "2cba9e42aa026125f3ed06c6d98c1db82f7ca027",
      "engine_version": "0.30.1rc1.dev616+g1a001d584",
      "image_digest": "h200/dsv41-dense-scale-rescue@sha256:d7bda81f69f56a15859bbc4ea3511a114383cb58c99b95e3257ec766418f7352",
      "layout": "trial_tp4_2rep",
      "gpus": "8",
      "max_model_len": "32768",
      "max_num_seqs": "32",
      "max_num_batched_tokens": "8192",
      "prefix_caching": "False",
      "spec_method": "",
      "spec_num_tokens": "",
      "workload": "random",
      "dataset": "",
      "concurrency": "16",
      "num_prompts": "64",
      "completed": "64",
      "failed": "0",
      "out_tok_s": "1340.7554204809137",
      "ttft_p50_ms": "401.5266740025254",
      "tpot_p50_ms": "10.37383669417799",
      "spec_accept_len": "",
      "spec_accept_rate": "",
      "prefix_hit_rate": ""
    },
    {
      "run_id": "dsv41_flash__trial_tp4_2rep__32k__vllm__native__1011-115241",
      "model": "dsv41_flash",
      "hf_repo": "deepseek-ai/DeepSeek-V4.1-Flash",
      "model_revision": "2cba9e42aa026125f3ed06c6d98c1db82f7ca027",
      "engine_version": "0.30.1rc1.dev616+g1a001d584",
      "image_digest": "h200/dsv41-dense-scale-rescue@sha256:d7bda81f69f56a15859bbc4ea3511a114383cb58c99b95e3257ec766418f7352",
      "layout": "trial_tp4_2rep",
      "gpus": "8",
      "max_model_len": "32768",
      "max_num_seqs": "32",
      "max_num_batched_tokens": "8192",
      "prefix_caching": "False",
      "spec_method": "",
      "spec_num_tokens": "",
      "workload": "random",
      "dataset": "",
      "concurrency": "64",
      "num_prompts": "128",
      "completed": "128",
      "failed": "0",
      "out_tok_s": "2788.0353353495666",
      "ttft_p50_ms": "857.3538924974855",
      "tpot_p50_ms": "19.09258721953294",
      "spec_accept_len": "",
      "spec_accept_rate": "",
      "prefix_hit_rate": ""
    }
  ],
  "limitations": [
    "Single pass per point; no confidence intervals.",
    "32K context configured; retrieval near31K.",
    "Real-chat uses8 prompts atc1 and32 different prompts atc16, max256 output, EOS honored.",
    "No combined two-replica DSpark run.",
    "Smoke checks do not establish quantization quality equivalence."
  ]
}