# MiniMax-M3: vLLM recipe

M3 serving recipe covering full and converted weights, hardware overrides and optional acceleration.

- Type: Deployment recipe
- Evidence: source_reported; site reproduced: false
- Model family: MiniMax
- Model: MiniMaxAI/MiniMax-M3
- Original author: vllm-project
- Original project: https://github.com/vllm-project/recipes
- Pinned document: https://github.com/vllm-project/recipes/blob/d5fc1f55cee70f4e1944f2dd4f7569cb8276ad94/models/MiniMaxAI/MiniMax-M3.yaml
- Source revision: d5fc1f55cee70f4e1944f2dd4f7569cb8276ad94
- Source commit time (not experiment date): 2026-10-17T09:22:45Z
- Checked: 2026-10-17
- Experiment date: None
- Runtime topics: vLLM
- Hardware topics: amd, nvidia
- Reported setup: {"system": null, "node_count": null, "parallelism": null, "kv_dtype": null, "context_limit_tokens": null, "experiment_date": null}
- Performance figures: attributed selections with per-point conditions; no cross-source ranking

## Conditions

Unknown is null. Recipe options are not measured benchmark settings.

```json
{
  "conditions": {
    "model_revision": null,
    "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
  },
  "test_profiles": [],
  "recipe_parameters": {
    "minimum_runtime_version": "0.24.0",
    "nightly_required": true,
    "model_context_limit_tokens": 1048576,
    "compatible_strategies": [
      "single_node_tp",
      "single_node_tep",
      "single_node_dep",
      "multi_node_tp",
      "multi_node_tp_pp",
      "multi_node_tep",
      "multi_node_dep",
      "pd_cluster"
    ],
    "hardware_claims": [
      {
        "profile": "h200",
        "maintainer_status": "verified"
      },
      {
        "profile": "b200",
        "maintainer_status": "verified"
      },
      {
        "profile": "b300",
        "maintainer_status": "verified"
      },
      {
        "profile": "mi300x",
        "maintainer_status": "verified"
      },
      {
        "profile": "mi355x",
        "maintainer_status": "verified"
      }
    ],
    "recipe_updated_on": "2026-10-06",
    "effective_default_minimum_runtime_version": "0.24.0"
  },
  "deployment_options": [
    {
      "variant": "default",
      "model_id": "MiniMaxAI/MiniMax-M3",
      "precision": "bf16",
      "supported_hardware": null,
      "model_revision": null,
      "declared_planning_minimum_gb": 1025,
      "minimum_runtime_version": "0.24.0",
      "extra_args": [],
      "tensor_parallel_by_hardware": null
    },
    {
      "variant": "mxfp8",
      "model_id": "MiniMaxAI/MiniMax-M3-MXFP8",
      "precision": "mxfp8",
      "supported_hardware": null,
      "model_revision": null,
      "declared_planning_minimum_gb": 513,
      "minimum_runtime_version": "0.24.0",
      "extra_args": [
        "--attention_config.indexer_kv_dtype",
        "fp8"
      ],
      "tensor_parallel_by_hardware": null
    },
    {
      "variant": "nvfp4",
      "model_id": "nvidia/MiniMax-M3-NVFP4",
      "precision": "nvfp4",
      "supported_hardware": null,
      "model_revision": null,
      "declared_planning_minimum_gb": 257,
      "minimum_runtime_version": "0.24.0",
      "extra_args": [
        "--trust-remote-code",
        "--kv-cache-dtype",
        "fp8",
        "--attention_config.backend",
        "FLASHINFER",
        "--attention_config.use_trtllm_attention",
        "true",
        "--attention_config.indexer_kv_dtype",
        "fp8",
        "--attention_config.minimax_m3_msa_decode_backend",
        "cutlass"
      ],
      "tensor_parallel_by_hardware": null
    },
    {
      "variant": "mxfp4",
      "model_id": "amd/MiniMax-M3-MXFP4",
      "precision": "mxfp4",
      "supported_hardware": [
        "mi355x"
      ],
      "model_revision": null,
      "declared_planning_minimum_gb": 257,
      "minimum_runtime_version": "0.24.0",
      "extra_args": [],
      "tensor_parallel_by_hardware": 8
    }
  ]
}
```

## Reading notes

- Recipe minimum engine version: 0.24.0.
- Default artifact: MiniMaxAI/MiniMax-M3.
- Variant and hardware settings are source declarations; select the matching original profile.

## Limits

- A generic recipe is not proof that a patched three-Spark report used identical settings.
- Input length, output policy, concurrency and cache are workload-specific; no single benchmark is inferred from the recipe.

## License declaration

{"declared_spdx": "Apache-2.0", "scope": "Recipe repository only; each model and conversion has independent distribution terms."}


## Editorial answer

The pinned MiniMax-M3 recipe defaults to MiniMaxAI/MiniMax-M3 (bf16). Its effective default-variant minimum is vLLM 0.24.0. Select the exact variant, hardware override and topology before treating it as a launch plan. No deployment was run by this site.

By Run AI Here editorial · 2026-10-17

AI-assisted source reading and original editorial synthesis; no independent model execution.

## Which artifact and engine does this recipe select?

The recipe identifies MiniMaxAI/MiniMax-M3 as its model family entry and selects MiniMaxAI/MiniMax-M3 by default. The default precision label is bf16; model-wide and variant-specific runtime requirements are kept separately below. The document revision fixes what the recipe said, while the weight revision and executed runtime revision remain separate fields. A model-family title alone cannot establish either artifact identity.

Sources: https://github.com/vllm-project/recipes/blob/d5fc1f55cee70f4e1944f2dd4f7569cb8276ad94/models/MiniMaxAI/MiniMax-M3.yaml#L127

## What needs special attention for this model?

The default is BF16 and is much larger than the three-Spark NVFP4 operator setup. The recipe’s 427B total / 26B active labels differ from the official card’s approximately 428B / 23B. Neither source is silently corrected here. The block-size setting, attention backend and DSpark choice travel with the selected precision and topology.

Sources: https://github.com/vllm-project/recipes/blob/d5fc1f55cee70f4e1944f2dd4f7569cb8276ad94/models/MiniMaxAI/MiniMax-M3.yaml#L21

## How should the settings be applied?

Choose one artifact variant and one hardware profile from the original recipe, then resolve its shared, variant, hardware, strategy and optional-feature settings together. The base arguments listed here are only one layer and are deliberately not presented as a complete runnable command. Check any nightly-image requirement and the model’s parsers before loading weights. A context declaration is a model limit, while batched-token and sequence settings control a serving workload.

Sources: https://github.com/vllm-project/recipes/blob/d5fc1f55cee70f4e1944f2dd4f7569cb8276ad94/models/MiniMaxAI/MiniMax-M3.yaml#L252

## What does the memory number mean?

The GB figure is the recipe’s declared planning minimum for the selected variant. It is not a site measurement, a universal peak allocation or a promise of maximum context. KV cache, auxiliary/draft models, image encoders, temporary workspace and active requests still require memory. Keep GB and GiB distinct, and use the actual artifact size and runtime allocation for a machine-specific budget.

Sources: https://github.com/vllm-project/recipes/blob/d5fc1f55cee70f4e1944f2dd4f7569cb8276ad94/models/MiniMaxAI/MiniMax-M3.yaml#L134

## What is needed to reproduce a deployment?

Save the model and conversion revisions, weight hashes, tokenizer/template, container or runtime digest, device count/topology and complete resolved launch. First check one short public request. For a subsequent measurement, fix input and output counts, concurrency, cache state, warmup, sampling, reasoning effort, repetitions and failures. The declarations on this page are source configuration facts; no throughput or quality outcome follows from them.

Sources: https://github.com/vllm-project/recipes/blob/d5fc1f55cee70f4e1944f2dd4f7569cb8276ad94/models/MiniMaxAI/MiniMax-M3.yaml#L290

## Attributed configuration facts

```json
[
  {
    "name": "Architecture",
    "value": "moe",
    "classification": "recipe_declaration",
    "evidence_url": "https://github.com/vllm-project/recipes/blob/d5fc1f55cee70f4e1944f2dd4f7569cb8276ad94/models/MiniMaxAI/MiniMax-M3.yaml#L34",
    "source_document_sha256": "1ffa05682199d5b354fa25116dfc19c8bbcdc782612b5a9eff774076b953c2fd"
  },
  {
    "name": "Total parameters as counted by recipe",
    "value": "427B",
    "classification": "recipe_declaration",
    "evidence_url": "https://github.com/vllm-project/recipes/blob/d5fc1f55cee70f4e1944f2dd4f7569cb8276ad94/models/MiniMaxAI/MiniMax-M3.yaml#L38",
    "source_document_sha256": "1ffa05682199d5b354fa25116dfc19c8bbcdc782612b5a9eff774076b953c2fd"
  },
  {
    "name": "Active parameters as counted by recipe",
    "value": "26B",
    "classification": "recipe_declaration",
    "evidence_url": "https://github.com/vllm-project/recipes/blob/d5fc1f55cee70f4e1944f2dd4f7569cb8276ad94/models/MiniMaxAI/MiniMax-M3.yaml#L39",
    "source_document_sha256": "1ffa05682199d5b354fa25116dfc19c8bbcdc782612b5a9eff774076b953c2fd"
  },
  {
    "name": "Native context declaration (tokens)",
    "value": 1048576,
    "classification": "recipe_declaration",
    "evidence_url": "https://github.com/vllm-project/recipes/blob/d5fc1f55cee70f4e1944f2dd4f7569cb8276ad94/models/MiniMaxAI/MiniMax-M3.yaml#L42",
    "source_document_sha256": "1ffa05682199d5b354fa25116dfc19c8bbcdc782612b5a9eff774076b953c2fd"
  },
  {
    "name": "Model-wide minimum vLLM",
    "value": "0.24.0",
    "classification": "recipe_declaration",
    "evidence_url": "https://github.com/vllm-project/recipes/blob/d5fc1f55cee70f4e1944f2dd4f7569cb8276ad94/models/MiniMaxAI/MiniMax-M3.yaml#L27",
    "source_document_sha256": "1ffa05682199d5b354fa25116dfc19c8bbcdc782612b5a9eff774076b953c2fd"
  },
  {
    "name": "Effective default-variant minimum vLLM",
    "value": "0.24.0",
    "classification": "recipe_declaration",
    "evidence_url": "https://github.com/vllm-project/recipes/blob/d5fc1f55cee70f4e1944f2dd4f7569cb8276ad94/models/MiniMaxAI/MiniMax-M3.yaml#L27",
    "source_document_sha256": "1ffa05682199d5b354fa25116dfc19c8bbcdc782612b5a9eff774076b953c2fd"
  },
  {
    "name": "Default artifact",
    "value": "MiniMaxAI/MiniMax-M3",
    "classification": "recipe_declaration",
    "evidence_url": "https://github.com/vllm-project/recipes/blob/d5fc1f55cee70f4e1944f2dd4f7569cb8276ad94/models/MiniMaxAI/MiniMax-M3.yaml#L127",
    "source_document_sha256": "1ffa05682199d5b354fa25116dfc19c8bbcdc782612b5a9eff774076b953c2fd"
  },
  {
    "name": "Default precision label",
    "value": "bf16",
    "classification": "recipe_declaration",
    "evidence_url": "https://github.com/vllm-project/recipes/blob/d5fc1f55cee70f4e1944f2dd4f7569cb8276ad94/models/MiniMaxAI/MiniMax-M3.yaml#L127",
    "source_document_sha256": "1ffa05682199d5b354fa25116dfc19c8bbcdc782612b5a9eff774076b953c2fd"
  },
  {
    "name": "Default planning minimum (GB; recipe declaration)",
    "value": 1025,
    "classification": "recipe_declaration",
    "evidence_url": "https://github.com/vllm-project/recipes/blob/d5fc1f55cee70f4e1944f2dd4f7569cb8276ad94/models/MiniMaxAI/MiniMax-M3.yaml#L134",
    "source_document_sha256": "1ffa05682199d5b354fa25116dfc19c8bbcdc782612b5a9eff774076b953c2fd"
  },
  {
    "name": "Default variant hardware allowlist",
    "value": null,
    "classification": "recipe_declaration",
    "evidence_url": "https://github.com/vllm-project/recipes/blob/d5fc1f55cee70f4e1944f2dd4f7569cb8276ad94/models/MiniMaxAI/MiniMax-M3.yaml#L127",
    "source_document_sha256": "1ffa05682199d5b354fa25116dfc19c8bbcdc782612b5a9eff774076b953c2fd"
  },
  {
    "name": "Shared base arguments (incomplete launch)",
    "value": [
      "--block-size",
      "128"
    ],
    "classification": "recipe_declaration",
    "evidence_url": "https://github.com/vllm-project/recipes/blob/d5fc1f55cee70f4e1944f2dd4f7569cb8276ad94/models/MiniMaxAI/MiniMax-M3.yaml#L127",
    "source_document_sha256": "1ffa05682199d5b354fa25116dfc19c8bbcdc782612b5a9eff774076b953c2fd"
  },
  {
    "name": "Declared compatible strategies",
    "value": [
      "single_node_tp",
      "single_node_tep",
      "single_node_dep",
      "multi_node_tp",
      "multi_node_tp_pp",
      "multi_node_tep",
      "multi_node_dep",
      "pd_cluster"
    ],
    "classification": "recipe_declaration",
    "evidence_url": "https://github.com/vllm-project/recipes/blob/d5fc1f55cee70f4e1944f2dd4f7569cb8276ad94/models/MiniMaxAI/MiniMax-M3.yaml#L127",
    "source_document_sha256": "1ffa05682199d5b354fa25116dfc19c8bbcdc782612b5a9eff774076b953c2fd"
  },
  {
    "name": "Optional feature names",
    "value": [
      "tool_calling",
      "reasoning",
      "spec_decoding",
      "text_only",
      "thinking_always_on",
      "encoder_parallel"
    ],
    "classification": "recipe_declaration",
    "evidence_url": "https://github.com/vllm-project/recipes/blob/d5fc1f55cee70f4e1944f2dd4f7569cb8276ad94/models/MiniMaxAI/MiniMax-M3.yaml#L127",
    "source_document_sha256": "1ffa05682199d5b354fa25116dfc19c8bbcdc782612b5a9eff774076b953c2fd"
  },
  {
    "name": "Features selected by default",
    "value": [
      "spec_decoding",
      "text_only",
      "thinking_always_on",
      "encoder_parallel"
    ],
    "classification": "recipe_declaration",
    "evidence_url": "https://github.com/vllm-project/recipes/blob/d5fc1f55cee70f4e1944f2dd4f7569cb8276ad94/models/MiniMaxAI/MiniMax-M3.yaml#L127",
    "source_document_sha256": "1ffa05682199d5b354fa25116dfc19c8bbcdc782612b5a9eff774076b953c2fd"
  }
]
```