{
 "id": "qwen2.5-vl-3b",
 "name": "Qwen2.5-VL 3B",
 "repo": "Qwen/Qwen2.5-VL-3B-Instruct",
 "lab": {
  "name": "Alibaba",
  "country": "China"
 },
 "released": "2025-01-26",
 "parameters": 3.754622976,
 "activeParameters": null,
 "maxContext": 128000,
 "architecture": "36 layers · 36 full attention (2 KV × 128)",
 "page": "https://studiotvai.com/vram-requirements/qwen2-5-vl-3b",
 "calculator": "https://studiotvai.com/llm-vram-calculator?model=qwen2.5-vl-3b",
 "weightsGiB": {
  "bf16": 6.99,
  "fp8": 3.79,
  "q4_k_m": 1.8
 },
 "kvCacheGiBPerRequest": [
  {
   "context": 8192,
   "bf16": 0.281,
   "fp8": 0.141
  },
  {
   "context": 32768,
   "bf16": 1.125,
   "fp8": 0.563
  }
 ],
 "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": 1,
     "fp8": 1,
     "q4_k_m": 1
    }
   },
   {
    "gpu": "rtx-5090",
    "name": "GeForce RTX 5090",
    "memoryGB": 32,
    "engine": "llama.cpp",
    "byFormat": {
     "bf16": 1,
     "fp8": 1,
     "q4_k_m": 1
    }
   },
   {
    "gpu": "rtx-pro-6000",
    "name": "RTX PRO 6000 Blackwell",
    "memoryGB": 96,
    "engine": "llama.cpp",
    "byFormat": {
     "bf16": 1,
     "fp8": 1,
     "q4_k_m": 1
    }
   },
   {
    "gpu": "mac-m5-max-128",
    "name": "Mac M5 Max 128 GB",
    "memoryGB": 96,
    "engine": "llama.cpp",
    "byFormat": {
     "bf16": 1,
     "fp8": 1,
     "q4_k_m": 1
    }
   },
   {
    "gpu": "h100",
    "name": "NVIDIA H100",
    "memoryGB": 80,
    "engine": "vLLM",
    "byFormat": {
     "bf16": 1,
     "fp8": 1,
     "q4_k_m": 1
    }
   },
   {
    "gpu": "h200",
    "name": "NVIDIA H200",
    "memoryGB": 141,
    "engine": "vLLM",
    "byFormat": {
     "bf16": 1,
     "fp8": 1,
     "q4_k_m": 1
    }
   },
   {
    "gpu": "b200",
    "name": "NVIDIA B200 (HGX)",
    "memoryGB": 180,
    "engine": "vLLM",
    "byFormat": {
     "bf16": 1,
     "fp8": 1,
     "q4_k_m": 1
    }
   }
  ]
 },
 "ggufFiles": {
  "repo": "unsloth/Qwen2.5-VL-3B-Instruct-GGUF",
  "gib": {
   "BF16": 5.75,
   "IQ4_NL": 1.7,
   "IQ4_XS": 1.62,
   "Q2_K": 1.19,
   "Q2_K_L": 1.19,
   "Q3_K_M": 1.48,
   "Q3_K_S": 1.35,
   "Q4_0": 1.7,
   "Q4_1": 1.86,
   "Q4_K_M": 1.8,
   "Q4_K_S": 1.71,
   "Q5_K_M": 2.07,
   "Q5_K_S": 2.02,
   "Q6_K": 2.36,
   "Q8_0": 3.06,
   "UD-IQ1_M": 0.82,
   "UD-IQ1_S": 0.77,
   "UD-IQ2_M": 1.09,
   "UD-IQ2_XXS": 0.9,
   "UD-IQ3_XXS": 1.21,
   "UD-Q2_K_XL": 1.22,
   "UD-Q3_K_XL": 1.52,
   "UD-Q4_K_XL": 1.82,
   "UD-Q5_K_XL": 2.08,
   "UD-Q6_K_XL": 2.58,
   "UD-Q8_K_XL": 3.66
  }
 },
 "runsWith": [
  "vLLM",
  "SGLang",
  "llama.cpp (Ollama, LM Studio)",
  "MLX (Mac)"
 ],
 "verified": {
  "gguf": {
   "day": "2026-10-14",
   "repo": "unsloth/Qwen2.5-VL-3B-Instruct-GGUF",
   "kv": "match",
   "ours": 36864,
   "theirs": 36864
  }
 },
 "updated": "2026-10-14T00:07:05.854Z"
}