{
  "checked_at": "2026-07-29T15:19:47.069938+00:00",
  "checkpoint": {
    "architecture": [
      "GlmMoeDsaForCausalLM"
    ],
    "bf16": {
      "individual_shard_names_archived": false,
      "last_modified": "2026-07-02T08:08:14.000Z",
      "revision": "b4734de4facf877f85769a911abafc5283eab3d9",
      "safetensor_bytes": 1506667387408,
      "safetensor_gb_decimal": 1506.667,
      "safetensor_shards": 282
    },
    "context_length": 1048576,
    "fp8": {
      "individual_shard_names_archived": false,
      "last_modified": "2026-07-02T08:09:27.000Z",
      "revision": "ba978f7d347eaf65d22f1a86833408afdb953541",
      "safetensor_bytes": 755632050320,
      "safetensor_gb_decimal": 755.632,
      "safetensor_shards": 141
    },
    "index_share_for_mtp_iteration": true,
    "model_type": "glm_moe_dsa",
    "nextn_layers": 1
  },
  "negative_controls": {
    "missing_nextn_layer": {
      "expected": "rejected",
      "observed": "rejected",
      "problems": [
        "checkpoint has no native next-token prediction layer"
      ]
    },
    "wrong_method": {
      "expected": "rejected",
      "observed": "rejected",
      "problems": [
        "built-in profile must use method=mtp"
      ]
    },
    "zero_draft_tokens": {
      "expected": "rejected",
      "observed": "rejected",
      "problems": [
        "num_speculative_tokens must be an integer from 1 to 5"
      ]
    }
  },
  "official_ablation_arithmetic": {
    "absolute_gain": 0.91,
    "baseline_accept_length": 4.56,
    "final_accept_length": 5.47,
    "not_equivalent_to_throughput_gain": true,
    "publisher_rounded_claim_percent": 20,
    "relative_gain_percent": 19.96
  },
  "runtime": {
    "image": "ghcr.io/astral-sh/uv@sha256:41977070f8f7a569ba48da4641bf5212421625e02763bbca24d6fe599ac62a91",
    "listener_opened": false,
    "network": "host network; approved outbound-only exception after the Snap Docker bridge DNS failure was reproduced",
    "platform": "Linux-5.15.0-139-generic-x86_64-with-glibc2.41",
    "published_port": false,
    "python": "3.13.13",
    "repository_mount": "read-only"
  },
  "schema_version": 1,
  "scope": {
    "downloaded_model_weights": false,
    "gpu_inference_run": false,
    "kind": "source-metadata-and-command preflight",
    "reason": "The host does not provide the multi-GPU HBM required by the 743B checkpoint; metadata validation cannot substitute for a serving benchmark.",
    "started_vllm_or_sglang_server": false,
    "throughput_benchmark_run": false
  },
  "secrets": {
    "credential_fields_read": false,
    "credential_required": false,
    "credential_values_archived": false
  },
  "sglang": {
    "low_latency_profile": {
      "algorithm": "EAGLE",
      "draft_tokens": 6,
      "steps": 5,
      "topk": 1
    },
    "minimum_version_from_glm_readme": "0.5.13.post1",
    "published_recipe_markers": {
      "eagle_algorithm": true,
      "five_steps": true,
      "one_nextn_layer": true,
      "one_topk": true,
      "six_draft_tokens": true,
      "tune_to_accept_length": true
    }
  },
  "source_receipts": [
    {
      "bytes": 3732,
      "content_type": "text/plain",
      "elapsed_ms": 810,
      "final_url": "https://huggingface.co/zai-org/GLM-5.2/raw/b4734de4facf877f85769a911abafc5283eab3d9/config.json",
      "name": "glm_config",
      "sha256": "185f93ee6d12548e16a847e279dc0c3c90b1524c970b0866b42fb545747d859a",
      "status": 200,
      "url": "https://huggingface.co/zai-org/GLM-5.2/raw/b4734de4facf877f85769a911abafc5283eab3d9/config.json"
    },
    {
      "bytes": 78171,
      "content_type": "application/json",
      "elapsed_ms": 1299,
      "final_url": "https://huggingface.co/api/models/zai-org/GLM-5.2/revision/b4734de4facf877f85769a911abafc5283eab3d9?blobs=true",
      "name": "glm_model_api",
      "sha256": "77b3ec86c0d824f7546a34107ccceade207575e1498d09649b84c0b0063e6ba3",
      "status": 200,
      "url": "https://huggingface.co/api/models/zai-org/GLM-5.2/revision/b4734de4facf877f85769a911abafc5283eab3d9?blobs=true"
    },
    {
      "bytes": 63284,
      "content_type": "application/json",
      "elapsed_ms": 832,
      "final_url": "https://huggingface.co/api/models/zai-org/GLM-5.2-FP8?blobs=true",
      "name": "glm_fp8_model_api",
      "sha256": "2d6cc05a06a846fed232ca3f7e718324da86fe11a6611f71a4dc2368efb66ff4",
      "status": 200,
      "url": "https://huggingface.co/api/models/zai-org/GLM-5.2-FP8?blobs=true"
    },
    {
      "bytes": 11849,
      "content_type": "text/plain",
      "elapsed_ms": 878,
      "final_url": "https://raw.githubusercontent.com/zai-org/GLM-5/436efa09bc868a6922e307624189e7018406beb9/README.md",
      "name": "glm_readme",
      "sha256": "fb229e4b461f6f5fbf60e8d79bb0e00a7ecd90996458a0888f00dd78e943ddc3",
      "status": 200,
      "url": "https://raw.githubusercontent.com/zai-org/GLM-5/436efa09bc868a6922e307624189e7018406beb9/README.md"
    },
    {
      "bytes": 19219,
      "content_type": "application/json",
      "elapsed_ms": 757,
      "final_url": "https://recipes.vllm.ai/zai-org/GLM-5.2.json",
      "name": "vllm_recipe_json",
      "sha256": "af7bdb69b3d761e038fc34af3cfab20e7b2c0330ea5fbcca06efc1780cf1c1d6",
      "status": 200,
      "url": "https://recipes.vllm.ai/zai-org/GLM-5.2.json"
    },
    {
      "bytes": 14918,
      "content_type": "text/plain",
      "elapsed_ms": 635,
      "final_url": "https://raw.githubusercontent.com/vllm-project/recipes/bdb35f4b88305153d1950e5e18e9074e633a76da/models/zai-org/GLM-5.2.yaml",
      "name": "vllm_recipe_yaml",
      "sha256": "481e0986f27cee0ae0094b73a1232198a07e3bff47b294767c22304e6686c19b",
      "status": 200,
      "url": "https://raw.githubusercontent.com/vllm-project/recipes/bdb35f4b88305153d1950e5e18e9074e633a76da/models/zai-org/GLM-5.2.yaml"
    },
    {
      "bytes": 197094,
      "content_type": "text/markdown",
      "elapsed_ms": 1187,
      "final_url": "https://docs.sglang.io/cookbook/autoregressive/GLM/GLM-5.2",
      "name": "sglang_recipe",
      "sha256": "a1d00015bc30a9ef068eeda81f5a1b00059213994cc2228283c0f2d1fa35ddbc",
      "status": 200,
      "url": "https://docs.sglang.io/cookbook/autoregressive/GLM/GLM-5.2"
    }
  ],
  "status": "pass",
  "vllm": {
    "bf16_vram_minimum_gb": 1786,
    "command_round_trip": "pass",
    "copyable_command": "vllm serve zai-org/GLM-5.2-FP8 --tensor-parallel-size 8 --kv-cache-dtype fp8 --speculative-config '{\"method\":\"mtp\",\"num_speculative_tokens\":5}' --tool-call-parser glm47 --reasoning-parser glm45 --enable-auto-tool-choice --served-model-name glm-5.2-fp8",
    "docker_image": "vllm/vllm-openai:v0.23.0",
    "fp8_vram_minimum_gb": 893,
    "minimum_version": "0.23.0",
    "mtp_config": {
      "method": "mtp",
      "num_speculative_tokens": 5
    },
    "pinned_yaml_markers": {
      "five_speculative_tokens": true,
      "method_mtp": true,
      "min_version_0_23_0": true
    },
    "profile_preflight": "pass",
    "tool_calling_and_mtp_note": "The current recipe says to use the latest main branch when tool calling and MTP are enabled together."
  }
}