{
 "id": "glm-5.3",
 "name": "GLM-5.3",
 "repo": "zai-org/GLM-5.3",
 "lab": {
  "name": "Z.ai",
  "country": "China"
 },
 "released": "2026-08-25",
 "parameters": 753.33,
 "activeParameters": 40.3,
 "maxContext": 1048576,
 "architecture": "MoE 256 experts · MLA (512 + 64 latent) on 78 layers · sparse-attention indexer keys on 21",
 "page": "https://studiotvai.com/vram-requirements/glm-5-3",
 "calculator": "https://studiotvai.com/llm-vram-calculator?model=glm-5.3",
 "weightsGiB": {
  "native": 694.39,
  "bf16": 1384.65,
  "fp8": 694.1,
  "q4_k_m": 418.18
 },
 "kvCacheGiBPerRequest": [
  {
   "context": 8192,
   "bf16": 0.707,
   "fp8": 0.412
  },
  {
   "context": 32768,
   "bf16": 2.827,
   "fp8": 1.646
  },
  {
   "context": 131072,
   "bf16": 11.307,
   "fp8": 6.584
  },
  {
   "context": 1048576,
   "bf16": 90.457,
   "fp8": 52.676
  }
 ],
 "recurrentStateGiBPerRequest": 0,
 "gpusNeeded": {
  "scenario": "one request of 8192 tokens, BF16 KV cache; vLLM on data-center GPUs, llama.cpp elsewhere",
  "gpus": [
   {
    "gpu": "rtx-4090",
    "name": "GeForce RTX 4090",
    "memoryGB": 24,
    "engine": "llama.cpp",
    "byFormat": {
     "native": 48,
     "bf16": 96,
     "fp8": 48,
     "q4_k_m": 24
    }
   },
   {
    "gpu": "rtx-5090",
    "name": "GeForce RTX 5090",
    "memoryGB": 32,
    "engine": "llama.cpp",
    "byFormat": {
     "native": 32,
     "bf16": 64,
     "fp8": 32,
     "q4_k_m": 16
    }
   },
   {
    "gpu": "rtx-pro-6000",
    "name": "RTX PRO 6000 Blackwell",
    "memoryGB": 96,
    "engine": "llama.cpp",
    "byFormat": {
     "native": 8,
     "bf16": 16,
     "fp8": 8,
     "q4_k_m": 5
    }
   },
   {
    "gpu": "mac-m5-max-128",
    "name": "Mac M5 Max 128 GB",
    "memoryGB": 96,
    "engine": "llama.cpp",
    "byFormat": {
     "native": 8,
     "bf16": 16,
     "fp8": 8,
     "q4_k_m": 5
    }
   },
   {
    "gpu": "h100",
    "name": "NVIDIA H100",
    "memoryGB": 80,
    "engine": "vLLM",
    "byFormat": {
     "native": 16,
     "bf16": 24,
     "fp8": 16,
     "q4_k_m": 7
    }
   },
   {
    "gpu": "h200",
    "name": "NVIDIA H200",
    "memoryGB": 141,
    "engine": "vLLM",
    "byFormat": {
     "native": 6,
     "bf16": 16,
     "fp8": 6,
     "q4_k_m": 4
    }
   },
   {
    "gpu": "b200",
    "name": "NVIDIA B200 (HGX)",
    "memoryGB": 180,
    "engine": "vLLM",
    "byFormat": {
     "native": 5,
     "bf16": 16,
     "fp8": 5,
     "q4_k_m": 3
    }
   }
  ]
 },
 "ggufFiles": {
  "repo": "unsloth/GLM-5.3-GGUF",
  "gib": {
   "BF16": 1404.42,
   "Q8_0": 746.32,
   "UD-IQ1_M": 212.8,
   "UD-IQ1_S": 201.83,
   "UD-IQ2_M": 222.19,
   "UD-IQ3_XXS": 262.34,
   "UD-IQ4_XS": 340.22,
   "UD-Q2_K_XL": 236.44,
   "UD-Q3_K_XL": 319.41,
   "UD-Q4_K_XL": 435.2,
   "UD-Q5_K_XL": 523.84,
   "UD-Q6_K_XL": 637.37
  }
 },
 "runsWith": [
  "vLLM",
  "SGLang",
  "llama.cpp (Ollama, LM Studio)",
  "MLX (Mac)"
 ],
 "verified": null,
 "updated": "2026-10-10T15:05:26.095Z"
}