{
  "id": "minimax-m3-three-sparks",
  "title": "MiniMax M3 on three DGX Sparks",
  "kind": "benchmark",
  "source_type": "community_operator_report",
  "summary": "An original operator repository describing a three-node TP=3 deployment and its benchmark checkpoint. It also retains a four-node baseline, so the two sections must be read as different conditions.",
  "runtime": "vLLM",
  "hardware_topics": [
    "spark"
  ],
  "platform_topics": [
    "linux"
  ],
  "model_label": "MiniMax M3 NVFP4 + DSpark",
  "source": {
    "publisher": "ChrisTheCoolHut",
    "url": "https://github.com/ChrisTheCoolHut/MiniMax-M3-NVFP4-DSpark-vLLM-3x-DGX-Spark",
    "repository": "ChrisTheCoolHut/MiniMax-M3-NVFP4-DSpark-vLLM-3x-DGX-Spark",
    "revision": "1008cf57355f49d5412f45880289516f29761f0b",
    "pinned_url": "https://github.com/ChrisTheCoolHut/MiniMax-M3-NVFP4-DSpark-vLLM-3x-DGX-Spark/tree/1008cf57355f49d5412f45880289516f29761f0b",
    "last_commit_at": "2026-08-05T23:05:16Z",
    "checked_on": "2026-10-17",
    "metadata_status": "checked",
    "report_url": "https://github.com/ChrisTheCoolHut/MiniMax-M3-NVFP4-DSpark-vLLM-3x-DGX-Spark/blob/1008cf57355f49d5412f45880289516f29761f0b/benchmarks/2026-08-05-tp3-dspark.md",
    "platform": "github",
    "document_path": "benchmarks/2026-08-05-tp3-dspark.md",
    "document_sha256": "fd06b6a7b2d74ead35c184ae7b31a99b7e43de3516a1dc5d56c1c218853682fc"
  },
  "license": {
    "declared_spdx": "MIT",
    "scope": "Repository declaration only; models, containers, datasets and per-file terms remain separate."
  },
  "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": null,
    "weight_hash": null,
    "runtime_execution_revision": null,
    "input_tokens": null,
    "output_tokens": null,
    "concurrency": null,
    "cache_state": null
  },
  "notes": [
    "The three-node path uses custom virtual-TP sharding patches.",
    "The repository distinguishes target weights, a DSpark drafter and FP8 KV cache.",
    "Read the dated three-node checkpoint and the actual recipe file together."
  ],
  "limitations": [
    "Three-node and four-node sections must not be merged.",
    "The checkpoint compares max_num_seqs=3 and 8; choose the matching experiment rather than treating them as one configuration.",
    "A declared container digest alone does not prove the TP=3 build matches the documented patch set."
  ],
  "comparison_status": "not_comparable",
  "editorial_status": "source_derived_detailed_article",
  "public_benchmark_snapshot": false,
  "reported_setup": {
    "system": "NVIDIA DGX Spark / GB10",
    "node_count": 3,
    "parallelism": "custom native vLLM TP=3",
    "kv_dtype": "FP8",
    "context_limit_tokens": 262144,
    "experiment_date": "2026-08-05"
  },
  "runtime_topics": [
    "vLLM"
  ],
  "model_family": "MiniMax",
  "model_id": null,
  "model_release_date": null,
  "model_artifact": null,
  "test_profiles": [],
  "deployment_options": [],
  "recipe_parameters": null,
  "measurement_scope": "inference_speed",
  "editorial": {
    "version": 1,
    "published_on": "2026-10-17",
    "modified_on": "2026-10-17",
    "byline": "Run AI Here editorial",
    "creation_method": "AI-assisted source reading and original editorial synthesis; no independent model execution.",
    "answer": "The operator pads attention-head layouts to make native tensor parallelism three possible, using a custom target and draft loader. The run uses three GB10 nodes linked in a switchless RoCE triangle, FP8 KV cache and a 262,144-token context ceiling. This is a patched serving stack rather than an ordinary unmodified vLLM launch. The report’s 80.41 GiB per-rank weights and 122-second load are source observations, not a capacity guarantee for another build.",
    "sections": [
      {
        "heading": "How does the three-node path work?",
        "paragraphs": [
          "The operator pads attention-head layouts to make native tensor parallelism three possible, using a custom target and draft loader. The run uses three GB10 nodes linked in a switchless RoCE triangle, FP8 KV cache and a 262,144-token context ceiling. This is a patched serving stack rather than an ordinary unmodified vLLM launch. The report’s 80.41 GiB per-rank weights and 122-second load are source observations, not a capacity guarantee for another build."
        ],
        "citations": [
          "https://github.com/ChrisTheCoolHut/MiniMax-M3-NVFP4-DSpark-vLLM-3x-DGX-Spark/blob/1008cf57355f49d5412f45880289516f29761f0b/benchmarks/2026-08-05-tp3-dspark.md#L9"
        ]
      },
      {
        "heading": "What changes when the sequence cap is raised?",
        "paragraphs": [
          "The greedy 80-item GSM8K sweep at eight parallel requests reports 64.63 aggregate tokens/s with max_num_seqs three and 97.28 with eight, while both report 98.8% accuracy. That comparison changes admission behavior. It should not be mixed with the short 128-token llama-benchy runs or interpreted as a 97-token/s single-user decode rate."
        ],
        "citations": [
          "https://github.com/ChrisTheCoolHut/MiniMax-M3-NVFP4-DSpark-vLLM-3x-DGX-Spark/blob/1008cf57355f49d5412f45880289516f29761f0b/benchmarks/2026-08-05-tp3-dspark.md#L102"
        ]
      },
      {
        "heading": "Why keep the sampled run separate?",
        "paragraphs": [
          "At the publisher’s temperature 1.0 and top_p 0.95, the parallel-eight quality run reports 97.5% accuracy and 86.46 aggregate tokens/s. Greedy and sampled runs answer different questions. The 80-item corpus is a small task-specific check; even a perfect result on that subset does not prove general equivalence of a patched sharding implementation."
        ],
        "citations": [
          "https://github.com/ChrisTheCoolHut/MiniMax-M3-NVFP4-DSpark-vLLM-3x-DGX-Spark/blob/1008cf57355f49d5412f45880289516f29761f0b/benchmarks/2026-08-05-tp3-dspark.md#L131"
        ]
      },
      {
        "heading": "What is still unknown?",
        "paragraphs": [
          "The record does not establish a full original weight hash or executed runtime commit. The document also compares older four-node data, which belongs to a separate configuration. Preserve the exact custom patch set, container, topology, cache policy and workload before attempting a reproduction. This site has not run the stack."
        ],
        "citations": [
          "https://github.com/ChrisTheCoolHut/MiniMax-M3-NVFP4-DSpark-vLLM-3x-DGX-Spark/blob/1008cf57355f49d5412f45880289516f29761f0b/benchmarks/2026-08-05-tp3-dspark.md#L80"
        ]
      }
    ],
    "facts": [],
    "reported_results": [
      {
        "label": "Short generation, depth 0, C1",
        "metric": "aggregate_output",
        "value": 23.55,
        "unit": "tokens/s",
        "statistic": "reported value",
        "uncertainty": null,
        "conditions": {
          "system": "NVIDIA DGX Spark / GB10",
          "node_count": 3,
          "parallelism": "custom native vLLM TP=3",
          "kv_dtype": "FP8",
          "context_limit_tokens": 262144,
          "experiment_date": "2026-08-05",
          "model": "nvidia/MiniMax-M3-NVFP4",
          "draft": "nvidia/MiniMax-M3-DSpark",
          "speculative_tokens": 8,
          "runtime": "custom vLLM TP3",
          "prefix_cache": true,
          "harness": "llama-benchy 0.4.0",
          "repetitions": 3,
          "max_num_seqs": 3,
          "output_tokens": 128,
          "retained_depth_tokens": 0,
          "concurrency": 1
        },
        "evidence_state": "source_reported",
        "site_reproduced": false,
        "evidence_url": "https://github.com/ChrisTheCoolHut/MiniMax-M3-NVFP4-DSpark-vLLM-3x-DGX-Spark/blob/1008cf57355f49d5412f45880289516f29761f0b/benchmarks/2026-08-05-tp3-dspark.md#L27",
        "source_document_sha256": "fd06b6a7b2d74ead35c184ae7b31a99b7e43de3516a1dc5d56c1c218853682fc"
      },
      {
        "label": "Generation, depth 100000, C1",
        "metric": "aggregate_output",
        "value": 26.83,
        "unit": "tokens/s",
        "statistic": "reported value",
        "uncertainty": null,
        "conditions": {
          "system": "NVIDIA DGX Spark / GB10",
          "node_count": 3,
          "parallelism": "custom native vLLM TP=3",
          "kv_dtype": "FP8",
          "context_limit_tokens": 262144,
          "experiment_date": "2026-08-05",
          "model": "nvidia/MiniMax-M3-NVFP4",
          "draft": "nvidia/MiniMax-M3-DSpark",
          "speculative_tokens": 8,
          "runtime": "custom vLLM TP3",
          "prefix_cache": true,
          "harness": "llama-benchy 0.4.0",
          "repetitions": 3,
          "max_num_seqs": 3,
          "output_tokens": 128,
          "retained_depth_tokens": 100000,
          "concurrency": 1
        },
        "evidence_state": "source_reported",
        "site_reproduced": false,
        "evidence_url": "https://github.com/ChrisTheCoolHut/MiniMax-M3-NVFP4-DSpark-vLLM-3x-DGX-Spark/blob/1008cf57355f49d5412f45880289516f29761f0b/benchmarks/2026-08-05-tp3-dspark.md#L50",
        "source_document_sha256": "fd06b6a7b2d74ead35c184ae7b31a99b7e43de3516a1dc5d56c1c218853682fc"
      },
      {
        "label": "GSM8K greedy, parallel 8, sequence cap 3",
        "metric": "aggregate_output",
        "value": 64.63,
        "unit": "tokens/s",
        "statistic": "reported value",
        "uncertainty": null,
        "conditions": {
          "system": "NVIDIA DGX Spark / GB10",
          "node_count": 3,
          "parallelism": "custom native vLLM TP=3",
          "kv_dtype": "FP8",
          "context_limit_tokens": 262144,
          "experiment_date": "2026-08-05",
          "model": "nvidia/MiniMax-M3-NVFP4",
          "draft": "nvidia/MiniMax-M3-DSpark",
          "speculative_tokens": 8,
          "runtime": "custom vLLM TP3",
          "prefix_cache": true,
          "harness": "llama-benchy 0.4.0",
          "repetitions": null,
          "max_num_seqs": 3,
          "corpus_items": 80,
          "temperature": 0,
          "concurrency": 8
        },
        "evidence_state": "source_reported",
        "site_reproduced": false,
        "evidence_url": "https://github.com/ChrisTheCoolHut/MiniMax-M3-NVFP4-DSpark-vLLM-3x-DGX-Spark/blob/1008cf57355f49d5412f45880289516f29761f0b/benchmarks/2026-08-05-tp3-dspark.md#L117",
        "source_document_sha256": "fd06b6a7b2d74ead35c184ae7b31a99b7e43de3516a1dc5d56c1c218853682fc"
      },
      {
        "label": "GSM8K greedy, parallel 8, sequence cap 8",
        "metric": "aggregate_output",
        "value": 97.28,
        "unit": "tokens/s",
        "statistic": "reported value",
        "uncertainty": null,
        "conditions": {
          "system": "NVIDIA DGX Spark / GB10",
          "node_count": 3,
          "parallelism": "custom native vLLM TP=3",
          "kv_dtype": "FP8",
          "context_limit_tokens": 262144,
          "experiment_date": "2026-08-05",
          "model": "nvidia/MiniMax-M3-NVFP4",
          "draft": "nvidia/MiniMax-M3-DSpark",
          "speculative_tokens": 8,
          "runtime": "custom vLLM TP3",
          "prefix_cache": true,
          "harness": "llama-benchy 0.4.0",
          "repetitions": null,
          "max_num_seqs": 8,
          "corpus_items": 80,
          "temperature": 0,
          "concurrency": 8
        },
        "evidence_state": "source_reported",
        "site_reproduced": false,
        "evidence_url": "https://github.com/ChrisTheCoolHut/MiniMax-M3-NVFP4-DSpark-vLLM-3x-DGX-Spark/blob/1008cf57355f49d5412f45880289516f29761f0b/benchmarks/2026-08-05-tp3-dspark.md#L125",
        "source_document_sha256": "fd06b6a7b2d74ead35c184ae7b31a99b7e43de3516a1dc5d56c1c218853682fc"
      },
      {
        "label": "GSM8K sampled, parallel 8",
        "metric": "accuracy",
        "value": 97.5,
        "unit": "percent",
        "statistic": "reported value",
        "uncertainty": null,
        "conditions": {
          "system": "NVIDIA DGX Spark / GB10",
          "node_count": 3,
          "parallelism": "custom native vLLM TP=3",
          "kv_dtype": "FP8",
          "context_limit_tokens": 262144,
          "experiment_date": "2026-08-05",
          "model": "nvidia/MiniMax-M3-NVFP4",
          "draft": "nvidia/MiniMax-M3-DSpark",
          "speculative_tokens": 8,
          "runtime": "custom vLLM TP3",
          "prefix_cache": true,
          "harness": "llama-benchy 0.4.0",
          "repetitions": null,
          "max_num_seqs": 8,
          "corpus_items": 80,
          "temperature": 1,
          "top_p": 0.95,
          "concurrency": 8
        },
        "evidence_state": "source_reported",
        "site_reproduced": false,
        "evidence_url": "https://github.com/ChrisTheCoolHut/MiniMax-M3-NVFP4-DSpark-vLLM-3x-DGX-Spark/blob/1008cf57355f49d5412f45880289516f29761f0b/benchmarks/2026-08-05-tp3-dspark.md#L140",
        "source_document_sha256": "fd06b6a7b2d74ead35c184ae7b31a99b7e43de3516a1dc5d56c1c218853682fc"
      }
    ],
    "source_document_sha256": "fd06b6a7b2d74ead35c184ae7b31a99b7e43de3516a1dc5d56c1c218853682fc"
  }
}