{
 "id": "gemma-4-31b",
 "name": "Gemma 4 31B",
 "repo": "google/gemma-4-31B-it",
 "lab": {
  "name": "Google",
  "country": "United States"
 },
 "released": "2026-03-11",
 "parameters": 31.27,
 "activeParameters": null,
 "maxContext": 262144,
 "architecture": "50 sliding-window layers (1024 tokens, 16 KV × 256) · 10 global layers (4 KV × 512)",
 "page": "https://studiotvai.com/vram-requirements/gemma-4-31b",
 "calculator": "https://studiotvai.com/llm-vram-calculator?model=gemma-4-31b",
 "weightsGiB": {
  "bf16": 58.24,
  "fp8": 30.43,
  "q4_k_m": 17.54
 },
 "kvCacheGiBPerRequest": [
  {
   "context": 8192,
   "bf16": 1.406,
   "fp8": 0.703
  },
  {
   "context": 32768,
   "bf16": 3.281,
   "fp8": 1.641
  },
  {
   "context": 131072,
   "bf16": 10.781,
   "fp8": 5.391
  },
  {
   "context": 262144,
   "bf16": 20.781,
   "fp8": 10.391
  }
 ],
 "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": 3,
     "fp8": 2,
     "q4_k_m": 1
    }
   },
   {
    "gpu": "rtx-5090",
    "name": "GeForce RTX 5090",
    "memoryGB": 32,
    "engine": "llama.cpp",
    "byFormat": {
     "bf16": 3,
     "fp8": 2,
     "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-4-31B-it-GGUF",
  "gib": {
   "BF16": 57.2,
   "IQ4_NL": 16.1,
   "IQ4_XS": 15.25,
   "Q3_K_M": 13.72,
   "Q3_K_S": 12.3,
   "Q4_0": 16.15,
   "Q4_1": 17.81,
   "Q4_K_M": 17.07,
   "Q4_K_S": 16.2,
   "Q5_K_M": 20.17,
   "Q5_K_S": 19.67,
   "Q6_K": 23.47,
   "Q8_0": 30.39,
   "UD-IQ2_M": 10.01,
   "UD-IQ2_XXS": 7.95,
   "UD-IQ3_XXS": 11.02,
   "UD-Q2_K_XL": 10.97,
   "UD-Q3_K_XL": 14.32,
   "UD-Q4_K_XL": 17.53,
   "UD-Q5_K_XL": 20.39,
   "UD-Q6_K_XL": 25.63,
   "UD-Q8_K_XL": 32.61
  }
 },
 "runsWith": [
  "vLLM",
  "SGLang",
  "llama.cpp (Ollama, LM Studio)",
  "MLX (Mac)"
 ],
 "verified": null,
 "updated": "2026-10-10T15:05:26.095Z"
}