data/model_registry.json

[
  {
    "model_name": "gpt-4o",
    "provider": "openai",
    "input_cost_per_1k": 0.0025,
    "output_cost_per_1k": 0.010,
    "context_window": 128000,
    "reasoning_score": 0.92,
    "coding_score": 0.93,
    "math_score": 0.90,
    "instruction_following_score": 0.95,
    "latency_score": 0.70,
    "max_complexity": 1.0,
    "notes": "Flagship multimodal model. Best quality/cost for hard tasks."
  },
  {
    "model_name": "gpt-4o-mini",
    "provider": "openai",
    "input_cost_per_1k": 0.000150,
    "output_cost_per_1k": 0.000600,
    "context_window": 128000,
    "reasoning_score": 0.78,
    "coding_score": 0.78,
    "math_score": 0.72,
    "instruction_following_score": 0.85,
    "latency_score": 0.90,
    "max_complexity": 0.75,
    "notes": "Cost-efficient small model for lightweight tasks."
  },
  {
    "model_name": "gpt-3.5-turbo",
    "provider": "openai",
    "input_cost_per_1k": 0.0005,
    "output_cost_per_1k": 0.0015,
    "context_window": 16385,
    "reasoning_score": 0.62,
    "coding_score": 0.65,
    "math_score": 0.55,
    "instruction_following_score": 0.75,
    "latency_score": 0.92,
    "max_complexity": 0.60,
    "notes": "Legacy fast model. Good for simple chat tasks."
  },
  {
    "model_name": "claude-3-5-haiku-20241022",
    "provider": "anthropic",
    "input_cost_per_1k": 0.00080,
    "output_cost_per_1k": 0.00400,
    "context_window": 200000,
    "reasoning_score": 0.78,
    "coding_score": 0.80,
    "math_score": 0.75,
    "instruction_following_score": 0.85,
    "latency_score": 0.92,
    "max_complexity": 0.75,
    "notes": "Fast, affordable Anthropic model for everyday tasks."
  },
  {
    "model_name": "claude-3-5-sonnet-20241022",
    "provider": "anthropic",
    "input_cost_per_1k": 0.003,
    "output_cost_per_1k": 0.015,
    "context_window": 200000,
    "reasoning_score": 0.93,
    "coding_score": 0.95,
    "math_score": 0.88,
    "instruction_following_score": 0.96,
    "latency_score": 0.75,
    "max_complexity": 1.0,
    "notes": "Top-tier coding and reasoning model from Anthropic."
  },
  {
    "model_name": "claude-3-haiku-20240307",
    "provider": "anthropic",
    "input_cost_per_1k": 0.00025,
    "output_cost_per_1k": 0.00125,
    "context_window": 200000,
    "reasoning_score": 0.65,
    "coding_score": 0.65,
    "math_score": 0.60,
    "instruction_following_score": 0.75,
    "latency_score": 0.95,
    "max_complexity": 0.60,
    "notes": "Cheapest Anthropic model. Good for classification, summarization."
  },
  {
    "model_name": "gemini-1.5-flash",
    "provider": "google",
    "input_cost_per_1k": 0.000075,
    "output_cost_per_1k": 0.000300,
    "context_window": 1000000,
    "reasoning_score": 0.74,
    "coding_score": 0.74,
    "math_score": 0.70,
    "instruction_following_score": 0.78,
    "latency_score": 0.88,
    "max_complexity": 0.72,
    "notes": "Extremely cheap and fast. Long context support."
  },
  {
    "model_name": "gemini-1.5-pro",
    "provider": "google",
    "input_cost_per_1k": 0.00125,
    "output_cost_per_1k": 0.005,
    "context_window": 2000000,
    "reasoning_score": 0.88,
    "coding_score": 0.87,
    "math_score": 0.85,
    "instruction_following_score": 0.90,
    "latency_score": 0.72,
    "max_complexity": 0.95,
    "notes": "Massive context window. Great for long-doc analysis."
  },
  {
    "model_name": "mistral-small-latest",
    "provider": "mistral",
    "input_cost_per_1k": 0.001,
    "output_cost_per_1k": 0.003,
    "context_window": 32000,
    "reasoning_score": 0.68,
    "coding_score": 0.70,
    "math_score": 0.62,
    "instruction_following_score": 0.75,
    "latency_score": 0.88,
    "max_complexity": 0.65,
    "notes": "Cost-effective European model."
  },
  {
    "model_name": "mistral-large-latest",
    "provider": "mistral",
    "input_cost_per_1k": 0.003,
    "output_cost_per_1k": 0.009,
    "context_window": 128000,
    "reasoning_score": 0.85,
    "coding_score": 0.84,
    "math_score": 0.80,
    "instruction_following_score": 0.88,
    "latency_score": 0.75,
    "max_complexity": 0.90,
    "notes": "Strong European flagship model."
  },
  {
    "model_name": "llama3.2:3b",
    "provider": "ollama",
    "input_cost_per_1k": 0.0,
    "output_cost_per_1k": 0.0,
    "context_window": 128000,
    "reasoning_score": 0.50,
    "coding_score": 0.48,
    "math_score": 0.42,
    "instruction_following_score": 0.60,
    "latency_score": 0.95,
    "max_complexity": 0.45,
    "notes": "Free local model. Use for simple/private tasks."
  },
  {
    "model_name": "llama3.1:8b",
    "provider": "ollama",
    "input_cost_per_1k": 0.0,
    "output_cost_per_1k": 0.0,
    "context_window": 128000,
    "reasoning_score": 0.65,
    "coding_score": 0.64,
    "math_score": 0.58,
    "instruction_following_score": 0.72,
    "latency_score": 0.85,
    "max_complexity": 0.62,
    "notes": "Free local model with decent reasoning."
  },
  {
    "model_name": "llama3.1:70b",
    "provider": "ollama",
    "input_cost_per_1k": 0.0,
    "output_cost_per_1k": 0.0,
    "context_window": 128000,
    "reasoning_score": 0.82,
    "coding_score": 0.82,
    "math_score": 0.78,
    "instruction_following_score": 0.85,
    "latency_score": 0.55,
    "max_complexity": 0.85,
    "notes": "Free local large model. Needs beefy hardware."
  },
  {
    "model_name": "deepseek-chat",
    "provider": "deepseek",
    "input_cost_per_1k": 0.00014,
    "output_cost_per_1k": 0.00028,
    "context_window": 64000,
    "reasoning_score": 0.88,
    "coding_score": 0.90,
    "math_score": 0.92,
    "instruction_following_score": 0.85,
    "latency_score": 0.72,
    "max_complexity": 0.92,
    "notes": "Exceptional value model especially strong in math and code."
  }
]