{
 "id": "qwen3-vl-235b-a22b",
 "name": "Qwen3-VL 235B-A22B",
 "repo": "Qwen/Qwen3-VL-235B-A22B-Instruct",
 "lab": {
  "name": "Alibaba",
  "country": "China"
 },
 "released": "2025-09-22",
 "parameters": 235.670022896,
 "activeParameters": 22,
 "maxContext": 262144,
 "architecture": "94 layers · 94 full attention (4 KV × 128)",
 "page": "https://studiotvai.com/vram-requirements/qwen3-vl-235b-a22b",
 "calculator": "https://studiotvai.com/llm-vram-calculator?model=qwen3-vl-235b-a22b",
 "weightsGiB": {
  "bf16": 438.97,
  "fp8": 220.64,
  "q4_k_m": 132.32
 },
 "kvCacheGiBPerRequest": [
  {
   "context": 8192,
   "bf16": 1.469,
   "fp8": 0.734
  },
  {
   "context": 32768,
   "bf16": 5.875,
   "fp8": 2.938
  },
  {
   "context": 131072,
   "bf16": 23.5,
   "fp8": 11.75
  },
  {
   "context": 262144,
   "bf16": 47,
   "fp8": 23.5
  }
 ],
 "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": {
     "bf16": 24,
     "fp8": 16,
     "q4_k_m": 7
    }
   },
   {
    "gpu": "rtx-5090",
    "name": "GeForce RTX 5090",
    "memoryGB": 32,
    "engine": "llama.cpp",
    "byFormat": {
     "bf16": 16,
     "fp8": 8,
     "q4_k_m": 5
    }
   },
   {
    "gpu": "rtx-pro-6000",
    "name": "RTX PRO 6000 Blackwell",
    "memoryGB": 96,
    "engine": "llama.cpp",
    "byFormat": {
     "bf16": 5,
     "fp8": 3,
     "q4_k_m": 2
    }
   },
   {
    "gpu": "mac-m5-max-128",
    "name": "Mac M5 Max 128 GB",
    "memoryGB": 96,
    "engine": "llama.cpp",
    "byFormat": {
     "bf16": 5,
     "fp8": 3,
     "q4_k_m": 2
    }
   },
   {
    "gpu": "strix-halo",
    "name": "Ryzen AI Max+ 395 128 GB",
    "memoryGB": 96,
    "engine": "llama.cpp",
    "byFormat": {
     "bf16": 5,
     "fp8": 3,
     "q4_k_m": 2
    }
   },
   {
    "gpu": "dgx-spark",
    "name": "DGX Spark (GB10)",
    "memoryGB": 128,
    "engine": "llama.cpp",
    "byFormat": {
     "bf16": 4,
     "fp8": 2,
     "q4_k_m": 2
    }
   },
   {
    "gpu": "h100",
    "name": "NVIDIA H100",
    "memoryGB": 80,
    "engine": "vLLM",
    "byFormat": {
     "bf16": 7,
     "fp8": 4,
     "q4_k_m": 2
    }
   },
   {
    "gpu": "h200",
    "name": "NVIDIA H200",
    "memoryGB": 141,
    "engine": "vLLM",
    "byFormat": {
     "bf16": 4,
     "fp8": 2,
     "q4_k_m": 2
    }
   },
   {
    "gpu": "b200",
    "name": "NVIDIA B200 (HGX)",
    "memoryGB": 180,
    "engine": "vLLM",
    "byFormat": {
     "bf16": 3,
     "fp8": 2,
     "q4_k_m": 1
    }
   }
  ]
 },
 "ggufFiles": {
  "repo": "unsloth/Qwen3-VL-235B-A22B-Instruct-GGUF",
  "gib": {
   "BF16": 437.99,
   "IQ4_NL": 123.5,
   "IQ4_XS": 116.7,
   "Q2_K": 79.81,
   "Q2_K_L": 79.94,
   "Q3_K_M": 104.72,
   "Q3_K_S": 94.48,
   "Q4_0": 123.99,
   "Q4_1": 137.12,
   "Q4_K_M": 132.39,
   "Q4_K_S": 124.51,
   "Q5_K_M": 155.36,
   "Q5_K_S": 150.76,
   "Q6_K": 179.76,
   "Q8_0": 232.77,
   "UD-TQ1_0": 50.67,
   "UD-IQ1_M": 64.83,
   "UD-IQ1_S": 58.49,
   "UD-IQ3_XXS": 88.15,
   "UD-Q2_K_XL": 81.97,
   "UD-Q3_K_XL": 96.6,
   "UD-Q4_K_XL": 124.91,
   "UD-Q5_K_XL": 155.45,
   "UD-Q6_K_XL": 186.65,
   "UD-Q8_K_XL": 251.16
  }
 },
 "runsWith": [
  "vLLM",
  "SGLang",
  "llama.cpp (Ollama, LM Studio)",
  "MLX (Mac)"
 ],
 "verified": {
  "gguf": {
   "day": "2026-11-06",
   "repo": "unsloth/Qwen3-VL-235B-A22B-Instruct-GGUF",
   "kv": "match",
   "ours": 192512,
   "theirs": 192512
  }
 },
 "updated": "2026-11-06T21:02:26.190Z"
}