{
 "id": "gemma-4-12b",
 "name": "Gemma 4 12B",
 "repo": "google/gemma-4-12B-it",
 "lab": {
  "name": "Google",
  "country": "United States"
 },
 "released": "2026-05-23",
 "parameters": 11.96,
 "activeParameters": null,
 "maxContext": 262144,
 "architecture": "40 sliding-window layers (1024 tokens, 8 KV × 256) · 8 global layers (1 KV × 512)",
 "page": "https://studiotvai.com/vram-requirements/gemma-4-12b",
 "calculator": "https://studiotvai.com/llm-vram-calculator?model=gemma-4-12b",
 "weightsGiB": {
  "bf16": 22.28,
  "fp8": 12.08,
  "q4_k_m": 6.9
 },
 "kvCacheGiBPerRequest": [
  {
   "context": 8192,
   "bf16": 0.438,
   "fp8": 0.219
  },
  {
   "context": 32768,
   "bf16": 0.813,
   "fp8": 0.406
  },
  {
   "context": 131072,
   "bf16": 2.313,
   "fp8": 1.156
  },
  {
   "context": 262144,
   "bf16": 4.313,
   "fp8": 2.156
  }
 ],
 "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": 2,
     "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/gemma-4-12b-it-GGUF",
  "gib": {
   "BF16": 22.2,
   "IQ4_NL": 6.26,
   "IQ4_XS": 5.94,
   "Q3_K_M": 5.3,
   "Q3_K_S": 4.78,
   "Q4_0": 6.28,
   "Q4_1": 6.89,
   "Q4_K_M": 6.63,
   "Q4_K_S": 6.3,
   "Q5_K_M": 7.84,
   "Q5_K_S": 7.64,
   "Q6_K": 9.11,
   "Q8_0": 11.8,
   "UD-IQ2_M": 3.92,
   "UD-IQ3_XXS": 4.32,
   "UD-Q2_K_XL": 4.34,
   "UD-Q3_K_XL": 5.61,
   "UD-Q4_K_XL": 6.86,
   "UD-Q5_K_XL": 8.01,
   "UD-Q6_K_XL": 9.95,
   "UD-Q8_K_XL": 12.7
  }
 },
 "runsWith": [
  "vLLM",
  "SGLang",
  "llama.cpp (Ollama, LM Studio)",
  "MLX (Mac)"
 ],
 "verified": null,
 "updated": "2026-10-10T15:05:26.095Z"
}