{
 "id": "lfm2-24b-a2b",
 "name": "LFM2 24B-A2B",
 "repo": "LiquidAI/LFM2-24B-A2B",
 "lab": {
  "name": "Liquid AI",
  "country": "United States"
 },
 "released": "2026-02-24",
 "parameters": 23.84,
 "activeParameters": 2.3,
 "maxContext": 128000,
 "architecture": "MoE 64 experts · 10 attention layers (GQA, 8 KV × 64) + 30 short-convolution layers",
 "page": "https://studiotvai.com/vram-requirements/lfm2-24b-a2b",
 "calculator": "https://studiotvai.com/llm-vram-calculator?model=lfm2-24b-a2b",
 "weightsGiB": {
  "bf16": 44.41,
  "fp8": 22.33,
  "q4_k_m": 13.43
 },
 "kvCacheGiBPerRequest": [
  {
   "context": 8192,
   "bf16": 0.156,
   "fp8": 0.078
  },
  {
   "context": 32768,
   "bf16": 0.625,
   "fp8": 0.313
  }
 ],
 "recurrentStateGiBPerRequest": 0.0003,
 "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": 2,
     "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": "bartowski/LiquidAI_LFM2-24B-A2B-GGUF",
  "gib": {
   "IQ2_M": 7.21,
   "IQ2_S": 6.45,
   "IQ2_XS": 6.17,
   "IQ2_XXS": 5.35,
   "IQ3_M": 10.09,
   "IQ3_XS": 9.11,
   "IQ3_XXS": 8.75,
   "IQ4_NL": 12.56,
   "IQ4_XS": 11.87,
   "Q2_K": 7.75,
   "Q2_K_L": 7.78,
   "Q3_K_L": 10.5,
   "Q3_K_M": 10.1,
   "Q3_K_S": 9.64,
   "Q3_K_XL": 10.53,
   "Q4_0": 12.77,
   "Q4_1": 13.93,
   "Q4_K_L": 13.47,
   "Q4_K_M": 13.44,
   "Q4_K_S": 12.99,
   "Q5_K_L": 15.8,
   "Q5_K_M": 15.77,
   "Q5_K_S": 15.31,
   "Q6_K": 18.24,
   "Q6_K_L": 18.27,
   "Q8_0": 23.61,
   "BF16": 44.42
  }
 },
 "runsWith": [
  "vLLM",
  "SGLang",
  "llama.cpp (Ollama, LM Studio)",
  "MLX (Mac)"
 ],
 "verified": {
  "gguf": {
   "day": "2026-10-10",
   "repo": "bartowski/LiquidAI_LFM2-24B-A2B-GGUF",
   "kv": "match",
   "ours": 20480,
   "theirs": 20480
  }
 },
 "updated": "2026-10-10T15:05:26.095Z"
}