{
  "id": "mimo26-flash-rl-official",
  "title": "MiMo-V2.6-Flash-RL: official model and deployment card",
  "kind": "model",
  "source_type": "official_model_card",
  "summary": "The RL checkpoint’s original card, with engine examples, parallelism and parser requirements. Keep this artifact separate from the MOPD variant.",
  "runtime": "vLLM / SGLang",
  "runtime_topics": [
    "vLLM",
    "SGLang"
  ],
  "hardware_topics": [],
  "platform_topics": [],
  "model_label": "MiMo-V2.6-Flash-RL",
  "model_family": "MiMo",
  "model_id": "XiaomiMiMo/MiMo-V2.6-Flash-RL",
  "model_release_date": null,
  "model_artifact": {
    "id": "XiaomiMiMo/MiMo-V2.6-Flash-RL",
    "revision": "dcd3fd83fb181622fbc606fbb1709901982a62c2",
    "format": null
  },
  "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": "dcd3fd83fb181622fbc606fbb1709901982a62c2",
    "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": null,
    "node_count": null,
    "parallelism": null,
    "kv_dtype": null,
    "context_limit_tokens": null,
    "experiment_date": null
  },
  "test_profiles": [],
  "deployment_options": [],
  "recipe_parameters": null,
  "measurement_scope": "official_capability_and_deployment",
  "notes": [
    "The SGLang and vLLM examples use different parallelism settings.",
    "The model-specific parsers and recommended sampling remain part of the configuration."
  ],
  "limitations": [
    "The related recipe defaults to MOPD; select the RL variant when reproducing this checkpoint."
  ],
  "comparison_status": "not_comparable",
  "editorial_status": "source_derived_detailed_article",
  "public_benchmark_snapshot": false,
  "source": {
    "publisher": "XiaomiMiMo",
    "url": "https://huggingface.co/XiaomiMiMo/MiMo-V2.6-Flash-RL",
    "repository": "XiaomiMiMo/MiMo-V2.6-Flash-RL",
    "platform": "huggingface",
    "revision": "dcd3fd83fb181622fbc606fbb1709901982a62c2",
    "document_path": "README.md",
    "pinned_url": "https://huggingface.co/XiaomiMiMo/MiMo-V2.6-Flash-RL/blob/dcd3fd83fb181622fbc606fbb1709901982a62c2/README.md",
    "report_url": null,
    "document_sha256": "132596ea76492320f5283756015e3f696f289b87fcf38e4b71602a8c901b0557",
    "last_commit_at": "2026-10-09T06:30:29.000Z",
    "checked_on": "2026-10-17",
    "metadata_status": "checked"
  },
  "license": {
    "declared_spdx": "mit",
    "declared_name": null,
    "scope": "Publisher model-card declaration; inspect the linked LICENSE and artifact-specific terms."
  },
  "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 RL checkpoint’s original card, with engine examples, parallelism and parser requirements. Keep this artifact separate from the MOPD variant.",
    "sections": [
      {
        "heading": "Which MiMo checkpoint is this?",
        "paragraphs": [
          "This is the RL checkpoint, not MOPD. Its card describes 309B total and 15B activated parameters, with a 1M-token context. The companion vLLM recipe now defaults to the MOPD artifact, so copying that default would change the weights. Always record the complete repository name and revision when following the recipe."
        ],
        "citations": [
          "https://huggingface.co/XiaomiMiMo/MiMo-V2.6-Flash-RL/blob/dcd3fd83fb181622fbc606fbb1709901982a62c2/README.md#L80"
        ]
      },
      {
        "heading": "Why do the two engine examples use different topology?",
        "paragraphs": [
          "The SGLang example uses tensor parallelism eight and data parallelism two, with DP attention and an EAGLE speculative path. The vLLM example uses tensor parallelism four and its own memory and parser settings. They are alternative launch examples; neither is a controlled comparison of two engines on identical hardware."
        ],
        "citations": [
          "https://huggingface.co/XiaomiMiMo/MiMo-V2.6-Flash-RL/blob/dcd3fd83fb181622fbc606fbb1709901982a62c2/README.md#L167"
        ]
      },
      {
        "heading": "Which model-specific controls belong in a record?",
        "paragraphs": [
          "Both serving paths select the mimo reasoning and tool-call parsers. The card recommends temperature 1.0 and top_p 0.95. The architecture table separates sliding-window and global-attention layers, and describes a multi-layer speculative decoder. A conventional full-attention KV estimate cannot simply be pasted over this hybrid design."
        ],
        "citations": [
          "https://huggingface.co/XiaomiMiMo/MiMo-V2.6-Flash-RL/blob/dcd3fd83fb181622fbc606fbb1709901982a62c2/README.md#L207"
        ]
      },
      {
        "heading": "How should the model-card benchmarks be interpreted?",
        "paragraphs": [
          "The results table labels a MiMo-V2.6 Flash column alongside other products; it does not establish speed on the topology of either local launch example. Keep the benchmark column identity, task protocol and actual checkpoint evidence separate. This record uses architecture and launch facts and does not turn those product-level quality cells into RL-specific workstation measurements."
        ],
        "citations": [
          "https://huggingface.co/XiaomiMiMo/MiMo-V2.6-Flash-RL/blob/dcd3fd83fb181622fbc606fbb1709901982a62c2/README.md#L98"
        ]
      }
    ],
    "facts": [
      {
        "name": "Total / active parameters",
        "value": "309B / 15B",
        "classification": "publisher_declaration",
        "evidence_url": "https://huggingface.co/XiaomiMiMo/MiMo-V2.6-Flash-RL/blob/dcd3fd83fb181622fbc606fbb1709901982a62c2/README.md#L80",
        "source_document_sha256": "132596ea76492320f5283756015e3f696f289b87fcf38e4b71602a8c901b0557"
      },
      {
        "name": "Published context",
        "value": "1M tokens",
        "classification": "publisher_declaration",
        "evidence_url": "https://huggingface.co/XiaomiMiMo/MiMo-V2.6-Flash-RL/blob/dcd3fd83fb181622fbc606fbb1709901982a62c2/README.md#L81",
        "source_document_sha256": "132596ea76492320f5283756015e3f696f289b87fcf38e4b71602a8c901b0557"
      },
      {
        "name": "SGLang TP / DP example",
        "value": "8 / 2",
        "classification": "publisher_declaration",
        "evidence_url": "https://huggingface.co/XiaomiMiMo/MiMo-V2.6-Flash-RL/blob/dcd3fd83fb181622fbc606fbb1709901982a62c2/README.md#L173",
        "source_document_sha256": "132596ea76492320f5283756015e3f696f289b87fcf38e4b71602a8c901b0557"
      },
      {
        "name": "vLLM TP example",
        "value": 4,
        "classification": "publisher_declaration",
        "evidence_url": "https://huggingface.co/XiaomiMiMo/MiMo-V2.6-Flash-RL/blob/dcd3fd83fb181622fbc606fbb1709901982a62c2/README.md#L197",
        "source_document_sha256": "132596ea76492320f5283756015e3f696f289b87fcf38e4b71602a8c901b0557"
      },
      {
        "name": "Recommended sampling",
        "value": "temperature 1.0; top_p 0.95",
        "classification": "publisher_declaration",
        "evidence_url": "https://huggingface.co/XiaomiMiMo/MiMo-V2.6-Flash-RL/blob/dcd3fd83fb181622fbc606fbb1709901982a62c2/README.md#L207",
        "source_document_sha256": "132596ea76492320f5283756015e3f696f289b87fcf38e4b71602a8c901b0557"
      }
    ],
    "reported_results": [],
    "source_document_sha256": "132596ea76492320f5283756015e3f696f289b87fcf38e4b71602a8c901b0557"
  }
}