{
  "id": "qwen38-flash-next-single-spark-context",
  "title": "Qwen Flash-Next: single-Spark memory and cold context",
  "kind": "benchmark",
  "source_type": "community_operator_report",
  "summary": "The v2.4.0 promotion arm serves the pinned GDN-MSE checkpoint, with an extra MXFP8 prefill shard. NVFP4 expert weights, MXFP8 paths and a refitted four-draft MTP head are separate components. The experimental image tag is retained; the later release-image digest does not prove the exact cold-test image digest. Repository release names are specific to this one-Spark recipe.",
  "runtime": "vLLM",
  "runtime_topics": [
    "vLLM"
  ],
  "hardware_topics": [
    "spark",
    "nvidia"
  ],
  "platform_topics": [
    "linux"
  ],
  "model_label": "Qwen3.8-Flash-Next",
  "model_family": "Qwen",
  "model_id": "ursuciprian/Qwen3.8-Flash-Next-NVFP4-GDN-MSE",
  "model_release_date": null,
  "model_artifact": {
    "id": "ursuciprian/Qwen3.8-Flash-Next-NVFP4-GDN-MSE",
    "revision": "03f4a0570496bbe741189f4a2a7d611679e85406"
  },
  "evidence_state": "source_reported",
  "site_reproduced": false,
  "record_type": "curated_reference",
  "performance_ranking_allowed": false,
  "metrics": null,
  "metrics_status": "site_metrics_unmeasured; attributed_source_figures_in_editorial.reported_results",
  "conditions": {
    "model_revision": "03f4a0570496bbe741189f4a2a7d611679e85406",
    "weight_hash": null,
    "runtime_execution_revision": null,
    "input_tokens": null,
    "output_tokens": null,
    "concurrency": null,
    "cache_state": null,
    "quantization": null,
    "speculative_decoding": null,
    "temperature": null,
    "top_p": null
  },
  "reported_setup": {
    "system": "one DGX Spark GB10 with 128 GB unified memory, local NVMe",
    "node_count": 1,
    "parallelism": "TP1",
    "kv_dtype": "fp8",
    "context_limit_tokens": 262144,
    "experiment_date": "2026-10-16"
  },
  "test_profiles": [
    {
      "label": "1x v2.4.0 16384-token ttft",
      "evidence_url": "https://github.com/ursuciprian/qwen3.8-flash-next-1x-dgx-spark/blob/348ede7e28a01f6aa6b501aaf2392d20a286a012/docs/data/capability.csv#L1262",
      "conditions": {
        "model": "Qwen3.8-Flash-Next",
        "model_artifact": "ursuciprian/Qwen3.8-Flash-Next-NVFP4-GDN-MSE",
        "model_revision": "03f4a0570496bbe741189f4a2a7d611679e85406",
        "extra_prefill_artifact": "local-inference-lab/Qwen3.8-Flash-Next-NVFP4",
        "extra_prefill_revision": "7c4f1bc1a2d6847e0cbc01ac6b823f00251de8dd",
        "weight_precision": "NVFP4 experts; MXFP8 dense/attention; GDN NVFP4 decode and MXFP8 prefill",
        "system": "one DGX Spark GB10 with 128 GB unified memory, local NVMe",
        "node_count": 1,
        "parallelism": "TP1",
        "engine": "vLLM + b12x",
        "engine_version": null,
        "runtime_execution_revision": null,
        "runtime_revision_prefix": "10d0acc8 files over 5dad364d",
        "b12x_revision_prefix": "a1ad1b69",
        "executed_image_tag": "ghcr.io/ursuciprian/spark-vllm-b12x:k103-20261015-a1ad1b69-10d0acc8",
        "executed_image_digest": null,
        "release_image_digest": "sha256:30fe678ed7ae5d4deb277be137497bcf597e46e829471b7bed24edd4294a84f1",
        "release_image_is_exact_cold_test_image": false,
        "recipe_release": "1x v2.4.0 promotion A/B; k103 cb test arm, not the release-image ship check",
        "experiment_date": "2026-10-16",
        "context_limit_tokens": 262144,
        "configured_max_requests": 8,
        "draft_tokens": 4,
        "kv_dtype": "fp8",
        "GPU_memory_utilization_startup_guard": 0.01,
        "configured_KV_pool_bytes": 15032385536,
        "PLE_backing": "NVMe mmap, page cache; reported 26.8 GiB PLE table",
        "concurrency": 1,
        "cache_state": "cold prefix, no cache hit; not a claimed cold OS page cache",
        "temperature": 0,
        "thinking": null,
        "output_tokens": 1,
        "actual_output_tokens_per_request": null,
        "GPU_clock_cap_MHz": null,
        "input_tokens": 16384,
        "repetitions": "4 boots x 3 cold prompts; SD between boots",
        "harness": "cold.py (k101)",
        "source_run_directory": "1x:results/k103-1x-prefill-combo-ab-20261016-1206/cb-01/",
        "source_metric": "ttft_s",
        "workload": "long-prompt"
      }
    },
    {
      "label": "1x v2.4.0 16384-token prefill_rate",
      "evidence_url": "https://github.com/ursuciprian/qwen3.8-flash-next-1x-dgx-spark/blob/348ede7e28a01f6aa6b501aaf2392d20a286a012/docs/data/capability.csv#L1263",
      "conditions": {
        "model": "Qwen3.8-Flash-Next",
        "model_artifact": "ursuciprian/Qwen3.8-Flash-Next-NVFP4-GDN-MSE",
        "model_revision": "03f4a0570496bbe741189f4a2a7d611679e85406",
        "extra_prefill_artifact": "local-inference-lab/Qwen3.8-Flash-Next-NVFP4",
        "extra_prefill_revision": "7c4f1bc1a2d6847e0cbc01ac6b823f00251de8dd",
        "weight_precision": "NVFP4 experts; MXFP8 dense/attention; GDN NVFP4 decode and MXFP8 prefill",
        "system": "one DGX Spark GB10 with 128 GB unified memory, local NVMe",
        "node_count": 1,
        "parallelism": "TP1",
        "engine": "vLLM + b12x",
        "engine_version": null,
        "runtime_execution_revision": null,
        "runtime_revision_prefix": "10d0acc8 files over 5dad364d",
        "b12x_revision_prefix": "a1ad1b69",
        "executed_image_tag": "ghcr.io/ursuciprian/spark-vllm-b12x:k103-20261015-a1ad1b69-10d0acc8",
        "executed_image_digest": null,
        "release_image_digest": "sha256:30fe678ed7ae5d4deb277be137497bcf597e46e829471b7bed24edd4294a84f1",
        "release_image_is_exact_cold_test_image": false,
        "recipe_release": "1x v2.4.0 promotion A/B; k103 cb test arm, not the release-image ship check",
        "experiment_date": "2026-10-16",
        "context_limit_tokens": 262144,
        "configured_max_requests": 8,
        "draft_tokens": 4,
        "kv_dtype": "fp8",
        "GPU_memory_utilization_startup_guard": 0.01,
        "configured_KV_pool_bytes": 15032385536,
        "PLE_backing": "NVMe mmap, page cache; reported 26.8 GiB PLE table",
        "concurrency": 1,
        "cache_state": "cold prefix, no cache hit; not a claimed cold OS page cache",
        "temperature": 0,
        "thinking": null,
        "output_tokens": 1,
        "actual_output_tokens_per_request": null,
        "GPU_clock_cap_MHz": null,
        "input_tokens": 16384,
        "repetitions": "4 boots x 3 cold prompts; SD between boots",
        "harness": "cold.py (k101)",
        "source_run_directory": "1x:results/k103-1x-prefill-combo-ab-20261016-1206/cb-01/",
        "source_metric": "pp",
        "workload": "long-prompt"
      }
    },
    {
      "label": "1x v2.4.0 65536-token ttft",
      "evidence_url": "https://github.com/ursuciprian/qwen3.8-flash-next-1x-dgx-spark/blob/348ede7e28a01f6aa6b501aaf2392d20a286a012/docs/data/capability.csv#L1264",
      "conditions": {
        "model": "Qwen3.8-Flash-Next",
        "model_artifact": "ursuciprian/Qwen3.8-Flash-Next-NVFP4-GDN-MSE",
        "model_revision": "03f4a0570496bbe741189f4a2a7d611679e85406",
        "extra_prefill_artifact": "local-inference-lab/Qwen3.8-Flash-Next-NVFP4",
        "extra_prefill_revision": "7c4f1bc1a2d6847e0cbc01ac6b823f00251de8dd",
        "weight_precision": "NVFP4 experts; MXFP8 dense/attention; GDN NVFP4 decode and MXFP8 prefill",
        "system": "one DGX Spark GB10 with 128 GB unified memory, local NVMe",
        "node_count": 1,
        "parallelism": "TP1",
        "engine": "vLLM + b12x",
        "engine_version": null,
        "runtime_execution_revision": null,
        "runtime_revision_prefix": "10d0acc8 files over 5dad364d",
        "b12x_revision_prefix": "a1ad1b69",
        "executed_image_tag": "ghcr.io/ursuciprian/spark-vllm-b12x:k103-20261015-a1ad1b69-10d0acc8",
        "executed_image_digest": null,
        "release_image_digest": "sha256:30fe678ed7ae5d4deb277be137497bcf597e46e829471b7bed24edd4294a84f1",
        "release_image_is_exact_cold_test_image": false,
        "recipe_release": "1x v2.4.0 promotion A/B; k103 cb test arm, not the release-image ship check",
        "experiment_date": "2026-10-16",
        "context_limit_tokens": 262144,
        "configured_max_requests": 8,
        "draft_tokens": 4,
        "kv_dtype": "fp8",
        "GPU_memory_utilization_startup_guard": 0.01,
        "configured_KV_pool_bytes": 15032385536,
        "PLE_backing": "NVMe mmap, page cache; reported 26.8 GiB PLE table",
        "concurrency": 1,
        "cache_state": "cold prefix, no cache hit; not a claimed cold OS page cache",
        "temperature": 0,
        "thinking": null,
        "output_tokens": 1,
        "actual_output_tokens_per_request": null,
        "GPU_clock_cap_MHz": null,
        "input_tokens": 65536,
        "repetitions": "4 boots x 3 cold prompts; SD between boots",
        "harness": "cold.py (k101)",
        "source_run_directory": "1x:results/k103-1x-prefill-combo-ab-20261016-1206/cb-01/",
        "source_metric": "ttft_s",
        "workload": "long-prompt"
      }
    },
    {
      "label": "1x v2.4.0 65536-token prefill_rate",
      "evidence_url": "https://github.com/ursuciprian/qwen3.8-flash-next-1x-dgx-spark/blob/348ede7e28a01f6aa6b501aaf2392d20a286a012/docs/data/capability.csv#L1265",
      "conditions": {
        "model": "Qwen3.8-Flash-Next",
        "model_artifact": "ursuciprian/Qwen3.8-Flash-Next-NVFP4-GDN-MSE",
        "model_revision": "03f4a0570496bbe741189f4a2a7d611679e85406",
        "extra_prefill_artifact": "local-inference-lab/Qwen3.8-Flash-Next-NVFP4",
        "extra_prefill_revision": "7c4f1bc1a2d6847e0cbc01ac6b823f00251de8dd",
        "weight_precision": "NVFP4 experts; MXFP8 dense/attention; GDN NVFP4 decode and MXFP8 prefill",
        "system": "one DGX Spark GB10 with 128 GB unified memory, local NVMe",
        "node_count": 1,
        "parallelism": "TP1",
        "engine": "vLLM + b12x",
        "engine_version": null,
        "runtime_execution_revision": null,
        "runtime_revision_prefix": "10d0acc8 files over 5dad364d",
        "b12x_revision_prefix": "a1ad1b69",
        "executed_image_tag": "ghcr.io/ursuciprian/spark-vllm-b12x:k103-20261015-a1ad1b69-10d0acc8",
        "executed_image_digest": null,
        "release_image_digest": "sha256:30fe678ed7ae5d4deb277be137497bcf597e46e829471b7bed24edd4294a84f1",
        "release_image_is_exact_cold_test_image": false,
        "recipe_release": "1x v2.4.0 promotion A/B; k103 cb test arm, not the release-image ship check",
        "experiment_date": "2026-10-16",
        "context_limit_tokens": 262144,
        "configured_max_requests": 8,
        "draft_tokens": 4,
        "kv_dtype": "fp8",
        "GPU_memory_utilization_startup_guard": 0.01,
        "configured_KV_pool_bytes": 15032385536,
        "PLE_backing": "NVMe mmap, page cache; reported 26.8 GiB PLE table",
        "concurrency": 1,
        "cache_state": "cold prefix, no cache hit; not a claimed cold OS page cache",
        "temperature": 0,
        "thinking": null,
        "output_tokens": 1,
        "actual_output_tokens_per_request": null,
        "GPU_clock_cap_MHz": null,
        "input_tokens": 65536,
        "repetitions": "4 boots x 3 cold prompts; SD between boots",
        "harness": "cold.py (k101)",
        "source_run_directory": "1x:results/k103-1x-prefill-combo-ab-20261016-1206/cb-01/",
        "source_metric": "pp",
        "workload": "long-prompt"
      }
    },
    {
      "label": "1x v2.4.0 131072-token ttft",
      "evidence_url": "https://github.com/ursuciprian/qwen3.8-flash-next-1x-dgx-spark/blob/348ede7e28a01f6aa6b501aaf2392d20a286a012/docs/data/capability.csv#L1266",
      "conditions": {
        "model": "Qwen3.8-Flash-Next",
        "model_artifact": "ursuciprian/Qwen3.8-Flash-Next-NVFP4-GDN-MSE",
        "model_revision": "03f4a0570496bbe741189f4a2a7d611679e85406",
        "extra_prefill_artifact": "local-inference-lab/Qwen3.8-Flash-Next-NVFP4",
        "extra_prefill_revision": "7c4f1bc1a2d6847e0cbc01ac6b823f00251de8dd",
        "weight_precision": "NVFP4 experts; MXFP8 dense/attention; GDN NVFP4 decode and MXFP8 prefill",
        "system": "one DGX Spark GB10 with 128 GB unified memory, local NVMe",
        "node_count": 1,
        "parallelism": "TP1",
        "engine": "vLLM + b12x",
        "engine_version": null,
        "runtime_execution_revision": null,
        "runtime_revision_prefix": "10d0acc8 files over 5dad364d",
        "b12x_revision_prefix": "a1ad1b69",
        "executed_image_tag": "ghcr.io/ursuciprian/spark-vllm-b12x:k103-20261015-a1ad1b69-10d0acc8",
        "executed_image_digest": null,
        "release_image_digest": "sha256:30fe678ed7ae5d4deb277be137497bcf597e46e829471b7bed24edd4294a84f1",
        "release_image_is_exact_cold_test_image": false,
        "recipe_release": "1x v2.4.0 promotion A/B; k103 cb test arm, not the release-image ship check",
        "experiment_date": "2026-10-16",
        "context_limit_tokens": 262144,
        "configured_max_requests": 8,
        "draft_tokens": 4,
        "kv_dtype": "fp8",
        "GPU_memory_utilization_startup_guard": 0.01,
        "configured_KV_pool_bytes": 15032385536,
        "PLE_backing": "NVMe mmap, page cache; reported 26.8 GiB PLE table",
        "concurrency": 1,
        "cache_state": "cold prefix, no cache hit; not a claimed cold OS page cache",
        "temperature": 0,
        "thinking": null,
        "output_tokens": 1,
        "actual_output_tokens_per_request": null,
        "GPU_clock_cap_MHz": null,
        "input_tokens": 131072,
        "repetitions": "4 boots x 3 cold prompts; SD between boots",
        "harness": "cold.py (k101)",
        "source_run_directory": "1x:results/k103-1x-prefill-combo-ab-20261016-1206/cb-01/",
        "source_metric": "ttft_s",
        "workload": "long-prompt"
      }
    },
    {
      "label": "1x v2.4.0 131072-token prefill_rate",
      "evidence_url": "https://github.com/ursuciprian/qwen3.8-flash-next-1x-dgx-spark/blob/348ede7e28a01f6aa6b501aaf2392d20a286a012/docs/data/capability.csv#L1267",
      "conditions": {
        "model": "Qwen3.8-Flash-Next",
        "model_artifact": "ursuciprian/Qwen3.8-Flash-Next-NVFP4-GDN-MSE",
        "model_revision": "03f4a0570496bbe741189f4a2a7d611679e85406",
        "extra_prefill_artifact": "local-inference-lab/Qwen3.8-Flash-Next-NVFP4",
        "extra_prefill_revision": "7c4f1bc1a2d6847e0cbc01ac6b823f00251de8dd",
        "weight_precision": "NVFP4 experts; MXFP8 dense/attention; GDN NVFP4 decode and MXFP8 prefill",
        "system": "one DGX Spark GB10 with 128 GB unified memory, local NVMe",
        "node_count": 1,
        "parallelism": "TP1",
        "engine": "vLLM + b12x",
        "engine_version": null,
        "runtime_execution_revision": null,
        "runtime_revision_prefix": "10d0acc8 files over 5dad364d",
        "b12x_revision_prefix": "a1ad1b69",
        "executed_image_tag": "ghcr.io/ursuciprian/spark-vllm-b12x:k103-20261015-a1ad1b69-10d0acc8",
        "executed_image_digest": null,
        "release_image_digest": "sha256:30fe678ed7ae5d4deb277be137497bcf597e46e829471b7bed24edd4294a84f1",
        "release_image_is_exact_cold_test_image": false,
        "recipe_release": "1x v2.4.0 promotion A/B; k103 cb test arm, not the release-image ship check",
        "experiment_date": "2026-10-16",
        "context_limit_tokens": 262144,
        "configured_max_requests": 8,
        "draft_tokens": 4,
        "kv_dtype": "fp8",
        "GPU_memory_utilization_startup_guard": 0.01,
        "configured_KV_pool_bytes": 15032385536,
        "PLE_backing": "NVMe mmap, page cache; reported 26.8 GiB PLE table",
        "concurrency": 1,
        "cache_state": "cold prefix, no cache hit; not a claimed cold OS page cache",
        "temperature": 0,
        "thinking": null,
        "output_tokens": 1,
        "actual_output_tokens_per_request": null,
        "GPU_clock_cap_MHz": null,
        "input_tokens": 131072,
        "repetitions": "4 boots x 3 cold prompts; SD between boots",
        "harness": "cold.py (k101)",
        "source_run_directory": "1x:results/k103-1x-prefill-combo-ab-20261016-1206/cb-01/",
        "source_metric": "pp",
        "workload": "long-prompt"
      }
    },
    {
      "label": "1x v2.4.0 allocated KV pool",
      "evidence_url": "https://github.com/ursuciprian/qwen3.8-flash-next-1x-dgx-spark/blob/348ede7e28a01f6aa6b501aaf2392d20a286a012/docs/BENCHMARKS.md#L106",
      "conditions": {
        "model": "Qwen3.8-Flash-Next",
        "model_artifact": "ursuciprian/Qwen3.8-Flash-Next-NVFP4-GDN-MSE",
        "model_revision": "03f4a0570496bbe741189f4a2a7d611679e85406",
        "extra_prefill_artifact": "local-inference-lab/Qwen3.8-Flash-Next-NVFP4",
        "extra_prefill_revision": "7c4f1bc1a2d6847e0cbc01ac6b823f00251de8dd",
        "weight_precision": "NVFP4 experts; MXFP8 dense/attention; GDN NVFP4 decode and MXFP8 prefill",
        "system": "one DGX Spark GB10 with 128 GB unified memory, local NVMe",
        "node_count": 1,
        "parallelism": "TP1",
        "engine": "vLLM + b12x",
        "engine_version": null,
        "runtime_execution_revision": null,
        "runtime_revision_prefix": "10d0acc8 files over 5dad364d",
        "b12x_revision_prefix": "a1ad1b69",
        "executed_image_tag": "ghcr.io/ursuciprian/spark-vllm-b12x:k103-20261015-a1ad1b69-10d0acc8",
        "executed_image_digest": null,
        "release_image_digest": "sha256:30fe678ed7ae5d4deb277be137497bcf597e46e829471b7bed24edd4294a84f1",
        "release_image_is_exact_cold_test_image": false,
        "recipe_release": "1x v2.4.0 promotion A/B; k103 cb test arm, not the release-image ship check",
        "experiment_date": "2026-10-16",
        "context_limit_tokens": 262144,
        "configured_max_requests": 8,
        "draft_tokens": 4,
        "kv_dtype": "fp8",
        "GPU_memory_utilization_startup_guard": 0.01,
        "configured_KV_pool_bytes": 15032385536,
        "PLE_backing": "NVMe mmap, page cache; reported 26.8 GiB PLE table",
        "concurrency": null,
        "cache_state": "cold prefix, no cache hit; not a claimed cold OS page cache",
        "temperature": null,
        "thinking": null,
        "output_tokens": null,
        "actual_output_tokens_per_request": null,
        "GPU_clock_cap_MHz": null,
        "input_tokens": null,
        "memory_kind": "allocated_KV_token_pool_not_measured_peak_RAM",
        "max_context_simultaneously_validated": false
      }
    }
  ],
  "deployment_options": [],
  "recipe_parameters": null,
  "measurement_scope": "source_memory_and_endpoint_observations",
  "notes": [
    "A source document, not an independent experiment count. Preserve measurement type and unknown fields."
  ],
  "limitations": [
    "No independent Run AI Here model execution.",
    "Document revision is distinct from execution and weight identity.",
    "File bytes, host RSS, anonymous memory, MemAvailable, KV allocation and configured context are different quantities."
  ],
  "comparison_status": "not_comparable",
  "editorial_status": "source_derived_detailed_article",
  "public_benchmark_snapshot": false,
  "source": {
    "publisher": "ursuciprian",
    "url": "https://github.com/ursuciprian/qwen3.8-flash-next-1x-dgx-spark",
    "repository": "ursuciprian/qwen3.8-flash-next-1x-dgx-spark",
    "platform": "github",
    "revision": "348ede7e28a01f6aa6b501aaf2392d20a286a012",
    "document_path": "README.md",
    "pinned_url": "https://github.com/ursuciprian/qwen3.8-flash-next-1x-dgx-spark/blob/348ede7e28a01f6aa6b501aaf2392d20a286a012/README.md",
    "report_url": "https://github.com/ursuciprian/qwen3.8-flash-next-1x-dgx-spark/blob/348ede7e28a01f6aa6b501aaf2392d20a286a012/README.md",
    "document_sha256": "bb618a879867f416c51cfc3fa8b699da427d28ec324b14af7d1279a2501ab37f",
    "last_commit_at": "2026-10-16T16:43:24Z",
    "checked_on": "2026-10-18",
    "metadata_status": "checked"
  },
  "supplemental_sources": [
    {
      "document_id": "qwen1x-docs-data-capability.csv",
      "parent_record_id": "qwen38-flash-next-single-spark-context",
      "publisher": "ursuciprian",
      "repository": "ursuciprian/qwen3.8-flash-next-1x-dgx-spark",
      "revision": "348ede7e28a01f6aa6b501aaf2392d20a286a012",
      "document_path": "docs/data/capability.csv",
      "pinned_url": "https://github.com/ursuciprian/qwen3.8-flash-next-1x-dgx-spark/blob/348ede7e28a01f6aa6b501aaf2392d20a286a012/docs/data/capability.csv",
      "document_sha256": "4376de2ea68ca23d5208731ce948a28801cd1da2dc026c9415b2b122e9339205",
      "checked_on": "2026-10-18"
    },
    {
      "document_id": "qwen1x-docs-BENCHMARKS.md",
      "parent_record_id": "qwen38-flash-next-single-spark-context",
      "publisher": "ursuciprian",
      "repository": "ursuciprian/qwen3.8-flash-next-1x-dgx-spark",
      "revision": "348ede7e28a01f6aa6b501aaf2392d20a286a012",
      "document_path": "docs/BENCHMARKS.md",
      "pinned_url": "https://github.com/ursuciprian/qwen3.8-flash-next-1x-dgx-spark/blob/348ede7e28a01f6aa6b501aaf2392d20a286a012/docs/BENCHMARKS.md",
      "document_sha256": "ac2128626b607b2e4571bdb45fb7e2639e5c65c83841a8f28367bb5692986fc0",
      "checked_on": "2026-10-18"
    },
    {
      "document_id": "qwen1x-VERSIONS.md",
      "parent_record_id": "qwen38-flash-next-single-spark-context",
      "publisher": "ursuciprian",
      "repository": "ursuciprian/qwen3.8-flash-next-1x-dgx-spark",
      "revision": "348ede7e28a01f6aa6b501aaf2392d20a286a012",
      "document_path": "VERSIONS.md",
      "pinned_url": "https://github.com/ursuciprian/qwen3.8-flash-next-1x-dgx-spark/blob/348ede7e28a01f6aa6b501aaf2392d20a286a012/VERSIONS.md",
      "document_sha256": "c513523883d95b716707697a0633fb9c5e63c9cf748aeaf37a38d4f771913abd",
      "checked_on": "2026-10-18"
    },
    {
      "document_id": "qwen1x-results-k103-1x-prefill-combo-ab-20261016-1206-cb-01-arm-p1-recipe.yaml",
      "parent_record_id": "qwen38-flash-next-single-spark-context",
      "publisher": "ursuciprian",
      "repository": "ursuciprian/qwen3.8-flash-next-1x-dgx-spark",
      "revision": "348ede7e28a01f6aa6b501aaf2392d20a286a012",
      "document_path": "results/k103-1x-prefill-combo-ab-20261016-1206/cb-01/arm-p1/recipe.yaml",
      "pinned_url": "https://github.com/ursuciprian/qwen3.8-flash-next-1x-dgx-spark/blob/348ede7e28a01f6aa6b501aaf2392d20a286a012/results/k103-1x-prefill-combo-ab-20261016-1206/cb-01/arm-p1/recipe.yaml",
      "document_sha256": "33ecbeea74e85987bd4dfa7418543438e98ff93ea2270c103deccab44a43fa7e",
      "checked_on": "2026-10-18"
    }
  ],
  "license": {
    "declared_spdx": "apache-2.0",
    "scope": "Source declaration only; base weights, converters, runtimes and source integration material retain their own terms."
  },
  "editorial": {
    "version": 1,
    "published_on": "2026-10-18",
    "modified_on": "2026-10-18",
    "byline": "Run AI Here editorial",
    "creation_method": "Original source-grounded editorial synthesis with local Spark Qwen assistance and Codex review; no site model benchmark.",
    "answer": "The v2.4.0 promotion arm serves the pinned GDN-MSE checkpoint, with an extra MXFP8 prefill shard. NVFP4 expert weights, MXFP8 paths and a refitted four-draft MTP head are separate components. The experimental image tag is retained; the later release-image digest does not prove the exact cold-test image digest. Repository release names are specific to this one-Spark recipe.",
    "sections": [
      {
        "heading": "Which quantization and build were tested?",
        "paragraphs": [
          "The v2.4.0 promotion arm serves the pinned GDN-MSE checkpoint, with an extra MXFP8 prefill shard. NVFP4 expert weights, MXFP8 paths and a refitted four-draft MTP head are separate components. The experimental image tag is retained; the later release-image digest does not prove the exact cold-test image digest. Repository release names are specific to this one-Spark recipe."
        ],
        "citations": [
          "https://github.com/ursuciprian/qwen3.8-flash-next-1x-dgx-spark/blob/348ede7e28a01f6aa6b501aaf2392d20a286a012/VERSIONS.md",
          "https://github.com/ursuciprian/qwen3.8-flash-next-1x-dgx-spark/blob/348ede7e28a01f6aa6b501aaf2392d20a286a012/results/k103-1x-prefill-combo-ab-20261016-1206/cb-01/arm-p1/recipe.yaml"
        ]
      },
      {
        "heading": "What does the long-prompt curve measure?",
        "paragraphs": [
          "The selected cold-prefix observations cover 16384, 65536 and 131072 input tokens at C1, temperature zero and one requested output token. Values are means of four boots with three cold prompts each; reported SD is between boot means. A cold prefix does not establish cold operating-system page cache. These are prefill/TTFT observations, not long-response decode or broad reasoning accuracy."
        ],
        "citations": [
          "https://github.com/ursuciprian/qwen3.8-flash-next-1x-dgx-spark/blob/348ede7e28a01f6aa6b501aaf2392d20a286a012/docs/data/capability.csv",
          "https://github.com/ursuciprian/qwen3.8-flash-next-1x-dgx-spark/blob/348ede7e28a01f6aa6b501aaf2392d20a286a012/docs/BENCHMARKS.md"
        ]
      },
      {
        "heading": "How should memory limits be read?",
        "paragraphs": [
          "The recipe allocates a 14 GiB FP8 KV pool and reports 993754 token capacity, while its request window is 262144 and scheduler ceiling is eight requests. The token pool is shared: these numbers do not demonstrate eight simultaneous full windows. PLE uses NVMe mmap and page cache. The 0.01 GPU-memory-utilization setting is a startup guard with explicit KV bytes, not measured one-percent memory use."
        ],
        "citations": [
          "https://github.com/ursuciprian/qwen3.8-flash-next-1x-dgx-spark/blob/348ede7e28a01f6aa6b501aaf2392d20a286a012/README.md",
          "https://github.com/ursuciprian/qwen3.8-flash-next-1x-dgx-spark/blob/348ede7e28a01f6aa6b501aaf2392d20a286a012/results/k103-1x-prefill-combo-ab-20261016-1206/cb-01/arm-p1/recipe.yaml"
        ]
      },
      {
        "heading": "How should two-Spark comparisons be selected?",
        "paragraphs": [
          "The project numbers releases independently for one-Spark and two-Spark setups. Its capability table also combines dated measurements when a newer build has no corresponding cell. Keep setup, release, workload and original run path together instead of treating every headline cell as one matched scaling experiment."
        ],
        "citations": [
          "https://github.com/ursuciprian/qwen3.8-flash-next-1x-dgx-spark/blob/348ede7e28a01f6aa6b501aaf2392d20a286a012/VERSIONS.md"
        ]
      }
    ],
    "facts": [],
    "reported_results": [
      {
        "label": "1x v2.4.0 16384-token ttft",
        "metric": "ttft",
        "value": 7.104,
        "unit": "s",
        "statistic": "mean of 4 boots x 3 cold prompts, sd between boots",
        "uncertainty": {
          "kind": "SD between boot means, not confidence interval",
          "standard_deviation": 0.016
        },
        "conditions": {
          "model": "Qwen3.8-Flash-Next",
          "model_artifact": "ursuciprian/Qwen3.8-Flash-Next-NVFP4-GDN-MSE",
          "model_revision": "03f4a0570496bbe741189f4a2a7d611679e85406",
          "extra_prefill_artifact": "local-inference-lab/Qwen3.8-Flash-Next-NVFP4",
          "extra_prefill_revision": "7c4f1bc1a2d6847e0cbc01ac6b823f00251de8dd",
          "weight_precision": "NVFP4 experts; MXFP8 dense/attention; GDN NVFP4 decode and MXFP8 prefill",
          "system": "one DGX Spark GB10 with 128 GB unified memory, local NVMe",
          "node_count": 1,
          "parallelism": "TP1",
          "engine": "vLLM + b12x",
          "engine_version": null,
          "runtime_execution_revision": null,
          "runtime_revision_prefix": "10d0acc8 files over 5dad364d",
          "b12x_revision_prefix": "a1ad1b69",
          "executed_image_tag": "ghcr.io/ursuciprian/spark-vllm-b12x:k103-20261015-a1ad1b69-10d0acc8",
          "executed_image_digest": null,
          "release_image_digest": "sha256:30fe678ed7ae5d4deb277be137497bcf597e46e829471b7bed24edd4294a84f1",
          "release_image_is_exact_cold_test_image": false,
          "recipe_release": "1x v2.4.0 promotion A/B; k103 cb test arm, not the release-image ship check",
          "experiment_date": "2026-10-16",
          "context_limit_tokens": 262144,
          "configured_max_requests": 8,
          "draft_tokens": 4,
          "kv_dtype": "fp8",
          "GPU_memory_utilization_startup_guard": 0.01,
          "configured_KV_pool_bytes": 15032385536,
          "PLE_backing": "NVMe mmap, page cache; reported 26.8 GiB PLE table",
          "concurrency": 1,
          "cache_state": "cold prefix, no cache hit; not a claimed cold OS page cache",
          "temperature": 0,
          "thinking": null,
          "output_tokens": 1,
          "actual_output_tokens_per_request": null,
          "GPU_clock_cap_MHz": null,
          "input_tokens": 16384,
          "repetitions": "4 boots x 3 cold prompts; SD between boots",
          "harness": "cold.py (k101)",
          "source_run_directory": "1x:results/k103-1x-prefill-combo-ab-20261016-1206/cb-01/",
          "source_metric": "ttft_s",
          "workload": "long-prompt"
        },
        "evidence_state": "source_reported",
        "site_reproduced": false,
        "source_document_id": "qwen1x-docs-data-capability.csv",
        "source_document_sha256": "4376de2ea68ca23d5208731ce948a28801cd1da2dc026c9415b2b122e9339205",
        "source_line": 1262,
        "evidence_url": "https://github.com/ursuciprian/qwen3.8-flash-next-1x-dgx-spark/blob/348ede7e28a01f6aa6b501aaf2392d20a286a012/docs/data/capability.csv#L1262"
      },
      {
        "label": "1x v2.4.0 16384-token prefill_rate",
        "metric": "prefill_rate",
        "value": 2306.4,
        "unit": "tokens/s",
        "statistic": "mean of 4 boots x 3 cold prompts, sd between boots",
        "uncertainty": {
          "kind": "SD between boot means, not confidence interval",
          "standard_deviation": 5.2
        },
        "conditions": {
          "model": "Qwen3.8-Flash-Next",
          "model_artifact": "ursuciprian/Qwen3.8-Flash-Next-NVFP4-GDN-MSE",
          "model_revision": "03f4a0570496bbe741189f4a2a7d611679e85406",
          "extra_prefill_artifact": "local-inference-lab/Qwen3.8-Flash-Next-NVFP4",
          "extra_prefill_revision": "7c4f1bc1a2d6847e0cbc01ac6b823f00251de8dd",
          "weight_precision": "NVFP4 experts; MXFP8 dense/attention; GDN NVFP4 decode and MXFP8 prefill",
          "system": "one DGX Spark GB10 with 128 GB unified memory, local NVMe",
          "node_count": 1,
          "parallelism": "TP1",
          "engine": "vLLM + b12x",
          "engine_version": null,
          "runtime_execution_revision": null,
          "runtime_revision_prefix": "10d0acc8 files over 5dad364d",
          "b12x_revision_prefix": "a1ad1b69",
          "executed_image_tag": "ghcr.io/ursuciprian/spark-vllm-b12x:k103-20261015-a1ad1b69-10d0acc8",
          "executed_image_digest": null,
          "release_image_digest": "sha256:30fe678ed7ae5d4deb277be137497bcf597e46e829471b7bed24edd4294a84f1",
          "release_image_is_exact_cold_test_image": false,
          "recipe_release": "1x v2.4.0 promotion A/B; k103 cb test arm, not the release-image ship check",
          "experiment_date": "2026-10-16",
          "context_limit_tokens": 262144,
          "configured_max_requests": 8,
          "draft_tokens": 4,
          "kv_dtype": "fp8",
          "GPU_memory_utilization_startup_guard": 0.01,
          "configured_KV_pool_bytes": 15032385536,
          "PLE_backing": "NVMe mmap, page cache; reported 26.8 GiB PLE table",
          "concurrency": 1,
          "cache_state": "cold prefix, no cache hit; not a claimed cold OS page cache",
          "temperature": 0,
          "thinking": null,
          "output_tokens": 1,
          "actual_output_tokens_per_request": null,
          "GPU_clock_cap_MHz": null,
          "input_tokens": 16384,
          "repetitions": "4 boots x 3 cold prompts; SD between boots",
          "harness": "cold.py (k101)",
          "source_run_directory": "1x:results/k103-1x-prefill-combo-ab-20261016-1206/cb-01/",
          "source_metric": "pp",
          "workload": "long-prompt"
        },
        "evidence_state": "source_reported",
        "site_reproduced": false,
        "source_document_id": "qwen1x-docs-data-capability.csv",
        "source_document_sha256": "4376de2ea68ca23d5208731ce948a28801cd1da2dc026c9415b2b122e9339205",
        "source_line": 1263,
        "evidence_url": "https://github.com/ursuciprian/qwen3.8-flash-next-1x-dgx-spark/blob/348ede7e28a01f6aa6b501aaf2392d20a286a012/docs/data/capability.csv#L1263"
      },
      {
        "label": "1x v2.4.0 65536-token ttft",
        "metric": "ttft",
        "value": 30.723,
        "unit": "s",
        "statistic": "mean of 4 boots x 3 cold prompts, sd between boots",
        "uncertainty": {
          "kind": "SD between boot means, not confidence interval",
          "standard_deviation": 0.06
        },
        "conditions": {
          "model": "Qwen3.8-Flash-Next",
          "model_artifact": "ursuciprian/Qwen3.8-Flash-Next-NVFP4-GDN-MSE",
          "model_revision": "03f4a0570496bbe741189f4a2a7d611679e85406",
          "extra_prefill_artifact": "local-inference-lab/Qwen3.8-Flash-Next-NVFP4",
          "extra_prefill_revision": "7c4f1bc1a2d6847e0cbc01ac6b823f00251de8dd",
          "weight_precision": "NVFP4 experts; MXFP8 dense/attention; GDN NVFP4 decode and MXFP8 prefill",
          "system": "one DGX Spark GB10 with 128 GB unified memory, local NVMe",
          "node_count": 1,
          "parallelism": "TP1",
          "engine": "vLLM + b12x",
          "engine_version": null,
          "runtime_execution_revision": null,
          "runtime_revision_prefix": "10d0acc8 files over 5dad364d",
          "b12x_revision_prefix": "a1ad1b69",
          "executed_image_tag": "ghcr.io/ursuciprian/spark-vllm-b12x:k103-20261015-a1ad1b69-10d0acc8",
          "executed_image_digest": null,
          "release_image_digest": "sha256:30fe678ed7ae5d4deb277be137497bcf597e46e829471b7bed24edd4294a84f1",
          "release_image_is_exact_cold_test_image": false,
          "recipe_release": "1x v2.4.0 promotion A/B; k103 cb test arm, not the release-image ship check",
          "experiment_date": "2026-10-16",
          "context_limit_tokens": 262144,
          "configured_max_requests": 8,
          "draft_tokens": 4,
          "kv_dtype": "fp8",
          "GPU_memory_utilization_startup_guard": 0.01,
          "configured_KV_pool_bytes": 15032385536,
          "PLE_backing": "NVMe mmap, page cache; reported 26.8 GiB PLE table",
          "concurrency": 1,
          "cache_state": "cold prefix, no cache hit; not a claimed cold OS page cache",
          "temperature": 0,
          "thinking": null,
          "output_tokens": 1,
          "actual_output_tokens_per_request": null,
          "GPU_clock_cap_MHz": null,
          "input_tokens": 65536,
          "repetitions": "4 boots x 3 cold prompts; SD between boots",
          "harness": "cold.py (k101)",
          "source_run_directory": "1x:results/k103-1x-prefill-combo-ab-20261016-1206/cb-01/",
          "source_metric": "ttft_s",
          "workload": "long-prompt"
        },
        "evidence_state": "source_reported",
        "site_reproduced": false,
        "source_document_id": "qwen1x-docs-data-capability.csv",
        "source_document_sha256": "4376de2ea68ca23d5208731ce948a28801cd1da2dc026c9415b2b122e9339205",
        "source_line": 1264,
        "evidence_url": "https://github.com/ursuciprian/qwen3.8-flash-next-1x-dgx-spark/blob/348ede7e28a01f6aa6b501aaf2392d20a286a012/docs/data/capability.csv#L1264"
      },
      {
        "label": "1x v2.4.0 65536-token prefill_rate",
        "metric": "prefill_rate",
        "value": 2133.2,
        "unit": "tokens/s",
        "statistic": "mean of 4 boots x 3 cold prompts, sd between boots",
        "uncertainty": {
          "kind": "SD between boot means, not confidence interval",
          "standard_deviation": 4.2
        },
        "conditions": {
          "model": "Qwen3.8-Flash-Next",
          "model_artifact": "ursuciprian/Qwen3.8-Flash-Next-NVFP4-GDN-MSE",
          "model_revision": "03f4a0570496bbe741189f4a2a7d611679e85406",
          "extra_prefill_artifact": "local-inference-lab/Qwen3.8-Flash-Next-NVFP4",
          "extra_prefill_revision": "7c4f1bc1a2d6847e0cbc01ac6b823f00251de8dd",
          "weight_precision": "NVFP4 experts; MXFP8 dense/attention; GDN NVFP4 decode and MXFP8 prefill",
          "system": "one DGX Spark GB10 with 128 GB unified memory, local NVMe",
          "node_count": 1,
          "parallelism": "TP1",
          "engine": "vLLM + b12x",
          "engine_version": null,
          "runtime_execution_revision": null,
          "runtime_revision_prefix": "10d0acc8 files over 5dad364d",
          "b12x_revision_prefix": "a1ad1b69",
          "executed_image_tag": "ghcr.io/ursuciprian/spark-vllm-b12x:k103-20261015-a1ad1b69-10d0acc8",
          "executed_image_digest": null,
          "release_image_digest": "sha256:30fe678ed7ae5d4deb277be137497bcf597e46e829471b7bed24edd4294a84f1",
          "release_image_is_exact_cold_test_image": false,
          "recipe_release": "1x v2.4.0 promotion A/B; k103 cb test arm, not the release-image ship check",
          "experiment_date": "2026-10-16",
          "context_limit_tokens": 262144,
          "configured_max_requests": 8,
          "draft_tokens": 4,
          "kv_dtype": "fp8",
          "GPU_memory_utilization_startup_guard": 0.01,
          "configured_KV_pool_bytes": 15032385536,
          "PLE_backing": "NVMe mmap, page cache; reported 26.8 GiB PLE table",
          "concurrency": 1,
          "cache_state": "cold prefix, no cache hit; not a claimed cold OS page cache",
          "temperature": 0,
          "thinking": null,
          "output_tokens": 1,
          "actual_output_tokens_per_request": null,
          "GPU_clock_cap_MHz": null,
          "input_tokens": 65536,
          "repetitions": "4 boots x 3 cold prompts; SD between boots",
          "harness": "cold.py (k101)",
          "source_run_directory": "1x:results/k103-1x-prefill-combo-ab-20261016-1206/cb-01/",
          "source_metric": "pp",
          "workload": "long-prompt"
        },
        "evidence_state": "source_reported",
        "site_reproduced": false,
        "source_document_id": "qwen1x-docs-data-capability.csv",
        "source_document_sha256": "4376de2ea68ca23d5208731ce948a28801cd1da2dc026c9415b2b122e9339205",
        "source_line": 1265,
        "evidence_url": "https://github.com/ursuciprian/qwen3.8-flash-next-1x-dgx-spark/blob/348ede7e28a01f6aa6b501aaf2392d20a286a012/docs/data/capability.csv#L1265"
      },
      {
        "label": "1x v2.4.0 131072-token ttft",
        "metric": "ttft",
        "value": 70.0,
        "unit": "s",
        "statistic": "mean of 4 boots x 3 cold prompts, sd between boots",
        "uncertainty": {
          "kind": "SD between boot means, not confidence interval",
          "standard_deviation": 0.137
        },
        "conditions": {
          "model": "Qwen3.8-Flash-Next",
          "model_artifact": "ursuciprian/Qwen3.8-Flash-Next-NVFP4-GDN-MSE",
          "model_revision": "03f4a0570496bbe741189f4a2a7d611679e85406",
          "extra_prefill_artifact": "local-inference-lab/Qwen3.8-Flash-Next-NVFP4",
          "extra_prefill_revision": "7c4f1bc1a2d6847e0cbc01ac6b823f00251de8dd",
          "weight_precision": "NVFP4 experts; MXFP8 dense/attention; GDN NVFP4 decode and MXFP8 prefill",
          "system": "one DGX Spark GB10 with 128 GB unified memory, local NVMe",
          "node_count": 1,
          "parallelism": "TP1",
          "engine": "vLLM + b12x",
          "engine_version": null,
          "runtime_execution_revision": null,
          "runtime_revision_prefix": "10d0acc8 files over 5dad364d",
          "b12x_revision_prefix": "a1ad1b69",
          "executed_image_tag": "ghcr.io/ursuciprian/spark-vllm-b12x:k103-20261015-a1ad1b69-10d0acc8",
          "executed_image_digest": null,
          "release_image_digest": "sha256:30fe678ed7ae5d4deb277be137497bcf597e46e829471b7bed24edd4294a84f1",
          "release_image_is_exact_cold_test_image": false,
          "recipe_release": "1x v2.4.0 promotion A/B; k103 cb test arm, not the release-image ship check",
          "experiment_date": "2026-10-16",
          "context_limit_tokens": 262144,
          "configured_max_requests": 8,
          "draft_tokens": 4,
          "kv_dtype": "fp8",
          "GPU_memory_utilization_startup_guard": 0.01,
          "configured_KV_pool_bytes": 15032385536,
          "PLE_backing": "NVMe mmap, page cache; reported 26.8 GiB PLE table",
          "concurrency": 1,
          "cache_state": "cold prefix, no cache hit; not a claimed cold OS page cache",
          "temperature": 0,
          "thinking": null,
          "output_tokens": 1,
          "actual_output_tokens_per_request": null,
          "GPU_clock_cap_MHz": null,
          "input_tokens": 131072,
          "repetitions": "4 boots x 3 cold prompts; SD between boots",
          "harness": "cold.py (k101)",
          "source_run_directory": "1x:results/k103-1x-prefill-combo-ab-20261016-1206/cb-01/",
          "source_metric": "ttft_s",
          "workload": "long-prompt"
        },
        "evidence_state": "source_reported",
        "site_reproduced": false,
        "source_document_id": "qwen1x-docs-data-capability.csv",
        "source_document_sha256": "4376de2ea68ca23d5208731ce948a28801cd1da2dc026c9415b2b122e9339205",
        "source_line": 1266,
        "evidence_url": "https://github.com/ursuciprian/qwen3.8-flash-next-1x-dgx-spark/blob/348ede7e28a01f6aa6b501aaf2392d20a286a012/docs/data/capability.csv#L1266"
      },
      {
        "label": "1x v2.4.0 131072-token prefill_rate",
        "metric": "prefill_rate",
        "value": 1872.5,
        "unit": "tokens/s",
        "statistic": "mean of 4 boots x 3 cold prompts, sd between boots",
        "uncertainty": {
          "kind": "SD between boot means, not confidence interval",
          "standard_deviation": 3.7
        },
        "conditions": {
          "model": "Qwen3.8-Flash-Next",
          "model_artifact": "ursuciprian/Qwen3.8-Flash-Next-NVFP4-GDN-MSE",
          "model_revision": "03f4a0570496bbe741189f4a2a7d611679e85406",
          "extra_prefill_artifact": "local-inference-lab/Qwen3.8-Flash-Next-NVFP4",
          "extra_prefill_revision": "7c4f1bc1a2d6847e0cbc01ac6b823f00251de8dd",
          "weight_precision": "NVFP4 experts; MXFP8 dense/attention; GDN NVFP4 decode and MXFP8 prefill",
          "system": "one DGX Spark GB10 with 128 GB unified memory, local NVMe",
          "node_count": 1,
          "parallelism": "TP1",
          "engine": "vLLM + b12x",
          "engine_version": null,
          "runtime_execution_revision": null,
          "runtime_revision_prefix": "10d0acc8 files over 5dad364d",
          "b12x_revision_prefix": "a1ad1b69",
          "executed_image_tag": "ghcr.io/ursuciprian/spark-vllm-b12x:k103-20261015-a1ad1b69-10d0acc8",
          "executed_image_digest": null,
          "release_image_digest": "sha256:30fe678ed7ae5d4deb277be137497bcf597e46e829471b7bed24edd4294a84f1",
          "release_image_is_exact_cold_test_image": false,
          "recipe_release": "1x v2.4.0 promotion A/B; k103 cb test arm, not the release-image ship check",
          "experiment_date": "2026-10-16",
          "context_limit_tokens": 262144,
          "configured_max_requests": 8,
          "draft_tokens": 4,
          "kv_dtype": "fp8",
          "GPU_memory_utilization_startup_guard": 0.01,
          "configured_KV_pool_bytes": 15032385536,
          "PLE_backing": "NVMe mmap, page cache; reported 26.8 GiB PLE table",
          "concurrency": 1,
          "cache_state": "cold prefix, no cache hit; not a claimed cold OS page cache",
          "temperature": 0,
          "thinking": null,
          "output_tokens": 1,
          "actual_output_tokens_per_request": null,
          "GPU_clock_cap_MHz": null,
          "input_tokens": 131072,
          "repetitions": "4 boots x 3 cold prompts; SD between boots",
          "harness": "cold.py (k101)",
          "source_run_directory": "1x:results/k103-1x-prefill-combo-ab-20261016-1206/cb-01/",
          "source_metric": "pp",
          "workload": "long-prompt"
        },
        "evidence_state": "source_reported",
        "site_reproduced": false,
        "source_document_id": "qwen1x-docs-data-capability.csv",
        "source_document_sha256": "4376de2ea68ca23d5208731ce948a28801cd1da2dc026c9415b2b122e9339205",
        "source_line": 1267,
        "evidence_url": "https://github.com/ursuciprian/qwen3.8-flash-next-1x-dgx-spark/blob/348ede7e28a01f6aa6b501aaf2392d20a286a012/docs/data/capability.csv#L1267"
      },
      {
        "label": "1x v2.4.0 allocated KV pool",
        "metric": "kv_pool_tokens",
        "value": 993754,
        "unit": "tokens",
        "statistic": "boot allocation observed on every reported boot",
        "uncertainty": null,
        "conditions": {
          "model": "Qwen3.8-Flash-Next",
          "model_artifact": "ursuciprian/Qwen3.8-Flash-Next-NVFP4-GDN-MSE",
          "model_revision": "03f4a0570496bbe741189f4a2a7d611679e85406",
          "extra_prefill_artifact": "local-inference-lab/Qwen3.8-Flash-Next-NVFP4",
          "extra_prefill_revision": "7c4f1bc1a2d6847e0cbc01ac6b823f00251de8dd",
          "weight_precision": "NVFP4 experts; MXFP8 dense/attention; GDN NVFP4 decode and MXFP8 prefill",
          "system": "one DGX Spark GB10 with 128 GB unified memory, local NVMe",
          "node_count": 1,
          "parallelism": "TP1",
          "engine": "vLLM + b12x",
          "engine_version": null,
          "runtime_execution_revision": null,
          "runtime_revision_prefix": "10d0acc8 files over 5dad364d",
          "b12x_revision_prefix": "a1ad1b69",
          "executed_image_tag": "ghcr.io/ursuciprian/spark-vllm-b12x:k103-20261015-a1ad1b69-10d0acc8",
          "executed_image_digest": null,
          "release_image_digest": "sha256:30fe678ed7ae5d4deb277be137497bcf597e46e829471b7bed24edd4294a84f1",
          "release_image_is_exact_cold_test_image": false,
          "recipe_release": "1x v2.4.0 promotion A/B; k103 cb test arm, not the release-image ship check",
          "experiment_date": "2026-10-16",
          "context_limit_tokens": 262144,
          "configured_max_requests": 8,
          "draft_tokens": 4,
          "kv_dtype": "fp8",
          "GPU_memory_utilization_startup_guard": 0.01,
          "configured_KV_pool_bytes": 15032385536,
          "PLE_backing": "NVMe mmap, page cache; reported 26.8 GiB PLE table",
          "concurrency": null,
          "cache_state": "cold prefix, no cache hit; not a claimed cold OS page cache",
          "temperature": null,
          "thinking": null,
          "output_tokens": null,
          "actual_output_tokens_per_request": null,
          "GPU_clock_cap_MHz": null,
          "input_tokens": null,
          "memory_kind": "allocated_KV_token_pool_not_measured_peak_RAM",
          "max_context_simultaneously_validated": false
        },
        "evidence_state": "source_reported",
        "site_reproduced": false,
        "source_document_id": "qwen1x-docs-BENCHMARKS.md",
        "source_document_sha256": "ac2128626b607b2e4571bdb45fb7e2639e5c65c83841a8f28367bb5692986fc0",
        "source_line": 106,
        "evidence_url": "https://github.com/ursuciprian/qwen3.8-flash-next-1x-dgx-spark/blob/348ede7e28a01f6aa6b501aaf2392d20a286a012/docs/BENCHMARKS.md#L106"
      }
    ],
    "source_document_sha256": "bb618a879867f416c51cfc3fa8b699da427d28ec324b14af7d1279a2501ab37f"
  },
  "memory_observations": [
    {
      "id": "qwen-1x-v2.4-cold",
      "model_artifact": "ursuciprian/Qwen3.8-Flash-Next-NVFP4-GDN-MSE",
      "checkpoint_revision": "03f4a0570496bbe741189f4a2a7d611679e85406",
      "extra_prefill_revision": "7c4f1bc1a2d6847e0cbc01ac6b823f00251de8dd",
      "engine": "vLLM + b12x",
      "image_tag": "ghcr.io/ursuciprian/spark-vllm-b12x:k103-20261015-a1ad1b69-10d0acc8",
      "executed_image_digest": null,
      "release_image_digest": "sha256:30fe678ed7ae5d4deb277be137497bcf597e46e829471b7bed24edd4294a84f1",
      "context_limit_tokens": 262144,
      "measured_cold_input_tokens": [
        16384,
        65536,
        131072
      ],
      "configured_max_requests": 8,
      "KV_pool_bytes": 15032385536,
      "KV_pool_tokens": 993754,
      "MTP_draft_tokens": 4,
      "PLE_file_reported_GiB": 26.8,
      "all_eight_full_window_requests_proven": false,
      "cold_prefix_is_not_cold_page_cache": true
    }
  ]
}