🌱 Inference Efficiency Analysis

{
  "service": "InferenceLatency.com",
  "endpoint": "/efficiency",
  "timestamp": "2025-09-07T23:20:22.536412Z",
  "efficiency_analysis": {
    "providers_analyzed": 7,
    "most_efficient_provider": "Google Gemini",
    "least_efficient_provider": "Groq",
    "average_energy_wh": 0.3486,
    "average_carbon_g": 0.1115,
    "grid_carbon_intensity": 320
  },
  "provider_efficiency": [
    {
      "provider": "Google Gemini",
      "model": "Gemini-2.0-Flash",
      "latency_ms": 276,
      "energy_wh_est": 0.24,
      "carbon_g_est": 0.0768,
      "carbon_per_1k_tokens": 76.8,
      "efficiency_score": 1.51,
      "grid_gco2_per_kwh": 320,
      "methodology": "estimated"
    },
    {
      "provider": "Together AI",
      "model": "Llama3.1-8B-Turbo",
      "latency_ms": 262,
      "energy_wh_est": 0.35,
      "carbon_g_est": 0.112,
      "carbon_per_1k_tokens": 112.0,
      "efficiency_score": 1.09,
      "grid_gco2_per_kwh": 320,
      "methodology": "estimated"
    },
    {
      "provider": "OpenAI",
      "model": "GPT-4o",
      "latency_ms": 281,
      "energy_wh_est": 0.4,
      "carbon_g_est": 0.128,
      "carbon_per_1k_tokens": 128.0,
      "efficiency_score": 0.89,
      "grid_gco2_per_kwh": 320,
      "methodology": "estimated"
    },
    {
      "provider": "Fireworks AI",
      "model": "Llama3.1-8B",
      "latency_ms": 368,
      "energy_wh_est": 0.35,
      "carbon_g_est": 0.112,
      "carbon_per_1k_tokens": 112.0,
      "efficiency_score": 0.78,
      "grid_gco2_per_kwh": 320,
      "methodology": "estimated"
    },
    {
      "provider": "OpenRouter",
      "model": "Mistral",
      "latency_ms": 759,
      "energy_wh_est": 0.35,
      "carbon_g_est": 0.112,
      "carbon_per_1k_tokens": 112.0,
      "efficiency_score": 0.38,
      "grid_gco2_per_kwh": 320,
      "methodology": "estimated"
    },
    {
      "provider": "Claude",
      "model": "Claude Sonnet 4",
      "latency_ms": 1366,
      "energy_wh_est": 0.4,
      "carbon_g_est": 0.128,
      "carbon_per_1k_tokens": 128.0,
      "efficiency_score": 0.18,
      "grid_gco2_per_kwh": 320,
      "methodology": "estimated"
    },
    {
      "provider": "Groq",
      "model": "llama-3.1-8b-instant",
      "latency_ms": 2008,
      "energy_wh_est": 0.35,
      "carbon_g_est": 0.112,
      "carbon_per_1k_tokens": 112.0,
      "efficiency_score": 0.14,
      "grid_gco2_per_kwh": 320,
      "methodology": "estimated"
    }
  ],
  "methodology": {
    "energy_source": "vendor_disclosures_and_estimates",
    "carbon_calculation": "(energy_wh / 1000) * grid_gco2_per_kwh",
    "grid_intensity_source": "global_average",
    "scope": "online_inference_only",
    "accuracy_note": "Estimates based on available vendor data and industry benchmarks"
  },
  "sustainability_insights": {
    "greenest_choice": "Google Gemini",
    "carbon_savings_potential": "Up to 35.2g CO₂e per 1k inferences",
    "recommendation": "Choose providers with lower energy consumption for high-volume inference workloads"
  }
}