{
 "id": "llama-4-scout",
 "name": "Llama 4 Scout 17B-16E",
 "repo": "meta-llama/Llama-4-Scout-17B-16E-Instruct",
 "lab": {
  "name": "Meta",
  "country": "United States"
 },
 "released": "2025-04-02",
 "parameters": 108.64,
 "activeParameters": 17,
 "maxContext": 10485760,
 "architecture": "MoE 16 experts · 36 chunked-attention layers (8192-token chunks, 8 KV × 128) · 12 global layers",
 "page": "https://studiotvai.com/vram-requirements/llama-4-scout",
 "calculator": "https://studiotvai.com/llm-vram-calculator?model=llama-4-scout",
 "weightsGiB": {
  "bf16": 202.36,
  "fp8": 103.11,
  "q4_k_m": 60.81
 },
 "kvCacheGiBPerRequest": [
  {
   "context": 8192,
   "bf16": 1.5,
   "fp8": 0.75
  },
  {
   "context": 32768,
   "bf16": 2.625,
   "fp8": 1.313
  },
  {
   "context": 131072,
   "bf16": 7.125,
   "fp8": 3.563
  },
  {
   "context": 10485760,
   "bf16": 481.125,
   "fp8": 240.563
  }
 ],
 "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": 16,
     "fp8": 6,
     "q4_k_m": 3
    }
   },
   {
    "gpu": "rtx-5090",
    "name": "GeForce RTX 5090",
    "memoryGB": 32,
    "engine": "llama.cpp",
    "byFormat": {
     "bf16": 8,
     "fp8": 4,
     "q4_k_m": 3
    }
   },
   {
    "gpu": "rtx-pro-6000",
    "name": "RTX PRO 6000 Blackwell",
    "memoryGB": 96,
    "engine": "llama.cpp",
    "byFormat": {
     "bf16": 3,
     "fp8": 2,
     "q4_k_m": 1
    }
   },
   {
    "gpu": "mac-m5-max-128",
    "name": "Mac M5 Max 128 GB",
    "memoryGB": 96,
    "engine": "llama.cpp",
    "byFormat": {
     "bf16": 3,
     "fp8": 2,
     "q4_k_m": 1
    }
   },
   {
    "gpu": "h100",
    "name": "NVIDIA H100",
    "memoryGB": 80,
    "engine": "vLLM",
    "byFormat": {
     "bf16": 3,
     "fp8": 2,
     "q4_k_m": 1
    }
   },
   {
    "gpu": "h200",
    "name": "NVIDIA H200",
    "memoryGB": 141,
    "engine": "vLLM",
    "byFormat": {
     "bf16": 2,
     "fp8": 1,
     "q4_k_m": 1
    }
   },
   {
    "gpu": "b200",
    "name": "NVIDIA B200 (HGX)",
    "memoryGB": 180,
    "engine": "vLLM",
    "byFormat": {
     "bf16": 2,
     "fp8": 1,
     "q4_k_m": 1
    }
   }
  ]
 },
 "ggufFiles": {
  "repo": "unsloth/Llama-4-Scout-17B-16E-Instruct-GGUF",
  "gib": {
   "BF16": 200.76,
   "IQ4_NL": 56.76,
   "IQ4_XS": 53.69,
   "Q2_K": 36.85,
   "Q2_K_L": 37.07,
   "Q3_K_S": 43.53,
   "UD-IQ1_M": 32.59,
   "UD-IQ1_S": 30.24,
   "UD-IQ2_M": 36.39,
   "UD-IQ2_XXS": 34.83,
   "UD-IQ3_XXS": 42.59,
   "UD-Q2_K_XL": 39.47,
   "UD-Q3_K_XL": 45.65,
   "UD-TQ1_0": 27.25,
   "Q3_K_M": 48.2,
   "Q4_0": 56.98,
   "Q4_1": 62.94,
   "Q4_K_M": 60.87,
   "Q4_K_S": 57.23,
   "Q5_K_M": 71.29,
   "Q5_K_S": 69.16,
   "Q6_K": 82.36,
   "Q8_0": 106.67,
   "UD-Q4_K_XL": 57.74,
   "UD-Q5_K_XL": 73.71,
   "UD-Q6_K_XL": 87.61,
   "UD-Q8_K_XL": 119.38
  }
 },
 "runsWith": [
  "vLLM",
  "SGLang",
  "llama.cpp (Ollama, LM Studio)",
  "MLX (Mac)"
 ],
 "verified": null,
 "updated": "2026-10-10T15:05:26.095Z"
}