{
 "id": "mistral-medium-3.5",
 "name": "Mistral Medium 3.5 128B",
 "repo": "mistralai/Mistral-Medium-3.5-128B",
 "lab": {
  "name": "Mistral AI",
  "country": "France"
 },
 "released": "2026-05-22",
 "parameters": 127.7,
 "activeParameters": null,
 "maxContext": 262144,
 "architecture": "88 layers · GQA, 8 KV heads × 128 · vision encoder",
 "page": "https://studiotvai.com/vram-requirements/mistral-medium-3-5",
 "calculator": "https://studiotvai.com/llm-vram-calculator?model=mistral-medium-3.5",
 "weightsGiB": {
  "native": 124.43,
  "bf16": 237.86,
  "fp8": 121.93,
  "q4_k_m": 70.62
 },
 "kvCacheGiBPerRequest": [
  {
   "context": 8192,
   "bf16": 2.75,
   "fp8": 1.375
  },
  {
   "context": 32768,
   "bf16": 11,
   "fp8": 5.5
  },
  {
   "context": 131072,
   "bf16": 44,
   "fp8": 22
  },
  {
   "context": 262144,
   "bf16": 88,
   "fp8": 44
  }
 ],
 "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": {
     "native": 7,
     "bf16": 16,
     "fp8": 7,
     "q4_k_m": 4
    }
   },
   {
    "gpu": "rtx-5090",
    "name": "GeForce RTX 5090",
    "memoryGB": 32,
    "engine": "llama.cpp",
    "byFormat": {
     "native": 5,
     "bf16": 16,
     "fp8": 5,
     "q4_k_m": 3
    }
   },
   {
    "gpu": "rtx-pro-6000",
    "name": "RTX PRO 6000 Blackwell",
    "memoryGB": 96,
    "engine": "llama.cpp",
    "byFormat": {
     "native": 2,
     "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": {
     "native": 2,
     "bf16": 3,
     "fp8": 2,
     "q4_k_m": 1
    }
   },
   {
    "gpu": "strix-halo",
    "name": "Ryzen AI Max+ 395 128 GB",
    "memoryGB": 96,
    "engine": "llama.cpp",
    "byFormat": {
     "native": 2,
     "bf16": 3,
     "fp8": 2,
     "q4_k_m": 1
    }
   },
   {
    "gpu": "dgx-spark",
    "name": "DGX Spark (GB10)",
    "memoryGB": 128,
    "engine": "llama.cpp",
    "byFormat": {
     "native": 1,
     "bf16": 2,
     "fp8": 1,
     "q4_k_m": 1
    }
   },
   {
    "gpu": "h100",
    "name": "NVIDIA H100",
    "memoryGB": 80,
    "engine": "vLLM",
    "byFormat": {
     "native": 2,
     "bf16": 4,
     "fp8": 2,
     "q4_k_m": 2
    }
   },
   {
    "gpu": "h200",
    "name": "NVIDIA H200",
    "memoryGB": 141,
    "engine": "vLLM",
    "byFormat": {
     "native": 2,
     "bf16": 2,
     "fp8": 2,
     "q4_k_m": 1
    }
   },
   {
    "gpu": "b200",
    "name": "NVIDIA B200 (HGX)",
    "memoryGB": 180,
    "engine": "vLLM",
    "byFormat": {
     "native": 1,
     "bf16": 2,
     "fp8": 1,
     "q4_k_m": 1
    }
   }
  ]
 },
 "ggufFiles": {
  "repo": "unsloth/Mistral-Medium-3.5-128B-GGUF",
  "gib": {
   "BF16": 232.89,
   "IQ4_NL": 66.03,
   "IQ4_XS": 62.51,
   "Q2_K": 43.39,
   "Q2_K_L": 43.74,
   "UD-IQ2_M": 41.08,
   "UD-IQ2_XXS": 32.54,
   "UD-IQ3_XXS": 45.87,
   "UD-Q2_K_XL": 44.82,
   "Q3_K_M": 56.46,
   "Q3_K_S": 50.63,
   "Q4_0": 66.12,
   "Q4_1": 73.08,
   "Q4_K_M": 69.75,
   "Q4_K_S": 66.36,
   "Q5_K_M": 82.25,
   "Q5_K_S": 80.27,
   "Q6_K": 95.53,
   "Q8_0": 123.73,
   "UD-Q3_K_XL": 58.21,
   "UD-Q4_K_XL": 70.49,
   "UD-Q5_K_XL": 82.31,
   "UD-Q6_K_XL": 101.66,
   "UD-Q8_K_XL": 134.8
  }
 },
 "runsWith": [
  "vLLM",
  "SGLang",
  "llama.cpp (Ollama, LM Studio)",
  "MLX (Mac)"
 ],
 "verified": {
  "gguf": {
   "day": "2026-10-30",
   "repo": "unsloth/Mistral-Medium-3.5-128B-GGUF",
   "kv": "match",
   "ours": 360448,
   "theirs": 360448
  }
 },
 "updated": "2026-10-30T00:07:05.915Z"
}