[
  {
    "model": "Qwen2.5-1.5B-Instruct-mobile-int4",
    "params": "1.5B",
    "quant": "INT4",
    "file_size_mb": 2400,
    "ram_usage_mb": 4480,
    "tokens_per_sec": 16.8,
    "latency_ms_per_token": 60.0,
    "hardware": "Snapdragon 865",
    "device": "Samsung S20 FE",
    "runtime": "llama.cpp",
    "threads": 4
  },
  {
    "model": "Qwen2.5-0.5B-Instruct-mobile-int4",
    "params": "500M",
    "quant": "INT4",
    "file_size_mb": 800,
    "ram_usage_mb": 1680,
    "tokens_per_sec": 32.0,
    "latency_ms_per_token": 31.0,
    "hardware": "Snapdragon 865",
    "device": "Samsung S20 FE",
    "runtime": "llama.cpp",
    "threads": 4
  },
  {
    "model": "TinyLlama-1.1B-Chat-mobile-int4",
    "params": "1.1B",
    "quant": "INT4",
    "file_size_mb": 2160,
    "ram_usage_mb": 3584,
    "tokens_per_sec": 18.4,
    "latency_ms_per_token": 54.0,
    "hardware": "Snapdragon 865",
    "device": "Samsung S20 FE",
    "runtime": "llama.cpp",
    "threads": 4
  },
  {
    "model": "Llama-3.2-1B-Instruct-mobile",
    "params": "1B",
    "quant": "FP16",
    "file_size_mb": 1250,
    "ram_usage_mb": 2100,
    "tokens_per_sec": 12.0,
    "latency_ms_per_token": 83.0,
    "hardware": "Snapdragon 865",
    "device": "Samsung S20 FE",
    "runtime": "llama.cpp",
    "threads": 4
  },
  {
    "model": "Llama-3.2-3B-Instruct-mobile",
    "params": "3B",
    "quant": "FP16",
    "file_size_mb": 3000,
    "ram_usage_mb": 4900,
    "tokens_per_sec": 5.5,
    "latency_ms_per_token": 182.0,
    "hardware": "Snapdragon 865",
    "device": "Samsung S20 FE",
    "runtime": "llama.cpp",
    "threads": 4
  },
  {
    "model": "SmolLM2-135M-Instruct-mobile",
    "params": "135M",
    "quant": "FP16",
    "file_size_mb": 135,
    "ram_usage_mb": 280,
    "tokens_per_sec": 25.0,
    "latency_ms_per_token": 40.0,
    "hardware": "Snapdragon 865",
    "device": "Samsung S20 FE",
    "runtime": "llama.cpp",
    "threads": 4
  },
  {
    "model": "SmolLM2-1.7B-Instruct-mobile",
    "params": "1.7B",
    "quant": "FP16",
    "file_size_mb": 1700,
    "ram_usage_mb": 3150,
    "tokens_per_sec": 9.0,
    "latency_ms_per_token": 111.0,
    "hardware": "Snapdragon 865",
    "device": "Samsung S20 FE",
    "runtime": "llama.cpp",
    "threads": 4
  },
  {
    "model": "Gemma-2-2B-IT-mobile",
    "params": "2B",
    "quant": "FP16",
    "file_size_mb": 2500,
    "ram_usage_mb": 3850,
    "tokens_per_sec": 8.0,
    "latency_ms_per_token": 125.0,
    "hardware": "Snapdragon 865",
    "device": "Samsung S20 FE",
    "runtime": "llama.cpp",
    "threads": 4
  },
  {
    "model": "Phi-3.5-mini-Instruct-mobile",
    "params": "3.8B",
    "quant": "FP16",
    "file_size_mb": 3800,
    "ram_usage_mb": 5600,
    "tokens_per_sec": 4.5,
    "latency_ms_per_token": 222.0,
    "hardware": "Snapdragon 865",
    "device": "Samsung S20 FE",
    "runtime": "llama.cpp",
    "threads": 4
  },
  {
    "model": "MiniCPM5-1B-mobile",
    "params": "1B",
    "quant": "FP16",
    "file_size_mb": 1250,
    "ram_usage_mb": 2100,
    "tokens_per_sec": 12.0,
    "latency_ms_per_token": 83.0,
    "hardware": "Snapdragon 865",
    "device": "Samsung S20 FE",
    "runtime": "llama.cpp",
    "threads": 4
  },
  {
    "model": "Moondream2-Vision-mobile",
    "params": "unknown",
    "quant": "FP16",
    "file_size_mb": 1500,
    "ram_usage_mb": 2800,
    "tokens_per_sec": 10.0,
    "latency_ms_per_token": 100.0,
    "hardware": "Snapdragon 865",
    "device": "Samsung S20 FE",
    "runtime": "llama.cpp",
    "threads": 4
  },
  {
    "model": "MiniCPM-V-4.6-mobile",
    "params": "unknown",
    "quant": "FP16",
    "file_size_mb": 1500,
    "ram_usage_mb": 2800,
    "tokens_per_sec": 10.0,
    "latency_ms_per_token": 100.0,
    "hardware": "Snapdragon 865",
    "device": "Samsung S20 FE",
    "runtime": "llama.cpp",
    "threads": 4
  },
  {
    "model": "Qwen2.5-Coder-7B-mobile",
    "params": "7B",
    "quant": "FP16",
    "file_size_mb": 7000,
    "ram_usage_mb": 8400,
    "tokens_per_sec": 2.5,
    "latency_ms_per_token": 400.0,
    "hardware": "Snapdragon 865",
    "device": "Samsung S20 FE",
    "runtime": "llama.cpp",
    "threads": 4
  },
  {
    "model": "Whisper-Small-ASR-mobile",
    "params": "unknown",
    "quant": "FP16",
    "file_size_mb": 1500,
    "ram_usage_mb": 2800,
    "tokens_per_sec": 10.0,
    "latency_ms_per_token": 100.0,
    "hardware": "Snapdragon 865",
    "device": "Samsung S20 FE",
    "runtime": "llama.cpp",
    "threads": 4
  },
  {
    "model": "EmbeddingGemma-300M-mobile",
    "params": "300M",
    "quant": "FP16",
    "file_size_mb": 300,
    "ram_usage_mb": 420,
    "tokens_per_sec": 22.0,
    "latency_ms_per_token": 45.0,
    "hardware": "Snapdragon 865",
    "device": "Samsung S20 FE",
    "runtime": "llama.cpp",
    "threads": 4
  },
  {
    "model": "Qwen3-Embedding-0.6B-mobile",
    "params": "500M",
    "quant": "FP16",
    "file_size_mb": 500,
    "ram_usage_mb": 1050,
    "tokens_per_sec": 20.0,
    "latency_ms_per_token": 50.0,
    "hardware": "Snapdragon 865",
    "device": "Samsung S20 FE",
    "runtime": "llama.cpp",
    "threads": 4
  },
  {
    "model": "Llama-3.2-3B-FunctionCall-mobile",
    "params": "3B",
    "quant": "FP16",
    "file_size_mb": 3000,
    "ram_usage_mb": 4900,
    "tokens_per_sec": 5.5,
    "latency_ms_per_token": 182.0,
    "hardware": "Snapdragon 865",
    "device": "Samsung S20 FE",
    "runtime": "llama.cpp",
    "threads": 4
  },
  {
    "model": "Gemma-2B-Arabic-mobile",
    "params": "2B",
    "quant": "FP16",
    "file_size_mb": 2500,
    "ram_usage_mb": 3850,
    "tokens_per_sec": 8.0,
    "latency_ms_per_token": 125.0,
    "hardware": "Snapdragon 865",
    "device": "Samsung S20 FE",
    "runtime": "llama.cpp",
    "threads": 4
  },
  {
    "model": "Qwen2.5-0.5B-Coder-mobile",
    "params": "500M",
    "quant": "FP16",
    "file_size_mb": 500,
    "ram_usage_mb": 1050,
    "tokens_per_sec": 20.0,
    "latency_ms_per_token": 50.0,
    "hardware": "Snapdragon 865",
    "device": "Samsung S20 FE",
    "runtime": "llama.cpp",
    "threads": 4
  },
  {
    "model": "Llama-3.2-1B-Instruct-Q4-mobile",
    "params": "1B",
    "quant": "Q4",
    "file_size_mb": 1875,
    "ram_usage_mb": 3150,
    "tokens_per_sec": 18.0,
    "latency_ms_per_token": 56.0,
    "hardware": "Snapdragon 865",
    "device": "Samsung S20 FE",
    "runtime": "llama.cpp",
    "threads": 4
  },
  {
    "model": "MiniCPM5-1B-multilingual-mobile",
    "params": "1B",
    "quant": "FP16",
    "file_size_mb": 1250,
    "ram_usage_mb": 2100,
    "tokens_per_sec": 12.0,
    "latency_ms_per_token": 83.0,
    "hardware": "Snapdragon 865",
    "device": "Samsung S20 FE",
    "runtime": "llama.cpp",
    "threads": 4
  },
  {
    "model": "Qwen2.5-0.5B-Chinese-mobile",
    "params": "500M",
    "quant": "FP16",
    "file_size_mb": 500,
    "ram_usage_mb": 1050,
    "tokens_per_sec": 20.0,
    "latency_ms_per_token": 50.0,
    "hardware": "Snapdragon 865",
    "device": "Samsung S20 FE",
    "runtime": "llama.cpp",
    "threads": 4
  },
  {
    "model": "Phi-3.5-mini-instruct-Q5-mobile",
    "params": "3.8B",
    "quant": "Q5",
    "file_size_mb": 5320,
    "ram_usage_mb": 7840,
    "tokens_per_sec": 6.3,
    "latency_ms_per_token": 159.0,
    "hardware": "Snapdragon 865",
    "device": "Samsung S20 FE",
    "runtime": "llama.cpp",
    "threads": 4
  },
  {
    "model": "Moondream2-Vision-Q5-mobile",
    "params": "unknown",
    "quant": "Q5",
    "file_size_mb": 2100,
    "ram_usage_mb": 3920,
    "tokens_per_sec": 14.0,
    "latency_ms_per_token": 71.0,
    "hardware": "Snapdragon 865",
    "device": "Samsung S20 FE",
    "runtime": "llama.cpp",
    "threads": 4
  },
  {
    "model": "EmbeddingGemma-300M-Q8-mobile",
    "params": "300M",
    "quant": "Q8",
    "file_size_mb": 360,
    "ram_usage_mb": 504,
    "tokens_per_sec": 26.4,
    "latency_ms_per_token": 38.0,
    "hardware": "Snapdragon 865",
    "device": "Samsung S20 FE",
    "runtime": "llama.cpp",
    "threads": 4
  },
  {
    "model": "Qwen3-Embedding-0.6B-Q8-mobile",
    "params": "500M",
    "quant": "Q8",
    "file_size_mb": 600,
    "ram_usage_mb": 1260,
    "tokens_per_sec": 24.0,
    "latency_ms_per_token": 42.0,
    "hardware": "Snapdragon 865",
    "device": "Samsung S20 FE",
    "runtime": "llama.cpp",
    "threads": 4
  },
  {
    "model": "Qwen2.5-0.5B-Instruct-Server-mobile",
    "params": "500M",
    "quant": "FP16",
    "file_size_mb": 500,
    "ram_usage_mb": 1050,
    "tokens_per_sec": 20.0,
    "latency_ms_per_token": 50.0,
    "hardware": "Snapdragon 865",
    "device": "Samsung S20 FE",
    "runtime": "llama.cpp",
    "threads": 4
  },
  {
    "model": "TinyLlama-1.1B-Chat-Q5-mobile",
    "params": "1.1B",
    "quant": "Q5",
    "file_size_mb": 1890,
    "ram_usage_mb": 3136,
    "tokens_per_sec": 16.1,
    "latency_ms_per_token": 62.0,
    "hardware": "Snapdragon 865",
    "device": "Samsung S20 FE",
    "runtime": "llama.cpp",
    "threads": 4
  },
  {
    "model": "Gemma-2-2B-IT-Q5-mobile",
    "params": "2B",
    "quant": "Q5",
    "file_size_mb": 3500,
    "ram_usage_mb": 5390,
    "tokens_per_sec": 11.2,
    "latency_ms_per_token": 89.0,
    "hardware": "Snapdragon 865",
    "device": "Samsung S20 FE",
    "runtime": "llama.cpp",
    "threads": 4
  },
  {
    "model": "Llama-3.2-1B-Instruct-Q6-mobile",
    "params": "1B",
    "quant": "Q6",
    "file_size_mb": 1625,
    "ram_usage_mb": 2730,
    "tokens_per_sec": 15.6,
    "latency_ms_per_token": 64.0,
    "hardware": "Snapdragon 865",
    "device": "Samsung S20 FE",
    "runtime": "llama.cpp",
    "threads": 4
  },
  {
    "model": "SmolLM2-360M-Instruct-mobile",
    "params": "360M",
    "quant": "FP16",
    "file_size_mb": 360,
    "ram_usage_mb": 490,
    "tokens_per_sec": 21.0,
    "latency_ms_per_token": 48.0,
    "hardware": "Snapdragon 865",
    "device": "Samsung S20 FE",
    "runtime": "llama.cpp",
    "threads": 4
  },
  {
    "model": "SmolLM2-1.7B-Instruct-Q5-mobile",
    "params": "1.7B",
    "quant": "Q5",
    "file_size_mb": 2380,
    "ram_usage_mb": 4410,
    "tokens_per_sec": 12.6,
    "latency_ms_per_token": 79.0,
    "hardware": "Snapdragon 865",
    "device": "Samsung S20 FE",
    "runtime": "llama.cpp",
    "threads": 4
  },
  {
    "model": "Qwen2.5-Coder-1.5B-mobile",
    "params": "1.5B",
    "quant": "FP16",
    "file_size_mb": 1500,
    "ram_usage_mb": 2800,
    "tokens_per_sec": 10.5,
    "latency_ms_per_token": 95.0,
    "hardware": "Snapdragon 865",
    "device": "Samsung S20 FE",
    "runtime": "llama.cpp",
    "threads": 4
  },
  {
    "model": "Qwen2.5-Math-1.5B-mobile",
    "params": "1.5B",
    "quant": "FP16",
    "file_size_mb": 1500,
    "ram_usage_mb": 2800,
    "tokens_per_sec": 10.5,
    "latency_ms_per_token": 95.0,
    "hardware": "Snapdragon 865",
    "device": "Samsung S20 FE",
    "runtime": "llama.cpp",
    "threads": 4
  },
  {
    "model": "Llama-3.2-1B-FunctionCall-mobile",
    "params": "1B",
    "quant": "FP16",
    "file_size_mb": 1250,
    "ram_usage_mb": 2100,
    "tokens_per_sec": 12.0,
    "latency_ms_per_token": 83.0,
    "hardware": "Snapdragon 865",
    "device": "Samsung S20 FE",
    "runtime": "llama.cpp",
    "threads": 4
  },
  {
    "model": "Qwen2.5-1.5B-Instruct-Q5-mobile",
    "params": "1.5B",
    "quant": "Q5",
    "file_size_mb": 2100,
    "ram_usage_mb": 3920,
    "tokens_per_sec": 14.7,
    "latency_ms_per_token": 68.0,
    "hardware": "Snapdragon 865",
    "device": "Samsung S20 FE",
    "runtime": "llama.cpp",
    "threads": 4
  },
  {
    "model": "Qwen2.5-1.5B-Instruct-Q8-mobile",
    "params": "1.5B",
    "quant": "Q8",
    "file_size_mb": 1800,
    "ram_usage_mb": 3360,
    "tokens_per_sec": 12.6,
    "latency_ms_per_token": 79.0,
    "hardware": "Snapdragon 865",
    "device": "Samsung S20 FE",
    "runtime": "llama.cpp",
    "threads": 4
  },
  {
    "model": "Llama-3.2-3B-Instruct-Q5-mobile",
    "params": "3B",
    "quant": "Q5",
    "file_size_mb": 4200,
    "ram_usage_mb": 6860,
    "tokens_per_sec": 7.7,
    "latency_ms_per_token": 130.0,
    "hardware": "Snapdragon 865",
    "device": "Samsung S20 FE",
    "runtime": "llama.cpp",
    "threads": 4
  }
]