benchlm.ai

{
  "type": "object",
  "properties": {
    "useCase": {
      "type": "string",
      "enum": [
        "coding",
        "agents",
        "multimodal",
        "writing",
        "analysis",
        "math",
        "general",
        "long-context",
        "tool-use",
        "web-research",
        "computer-use",
        "document-ai",
        "image-understanding",
        "frontend-app-dev",
        "factuality"
      ],
      "description": "Primary model use case."
    },
    "task": {
      "type": "string",
      "enum": [
        "emails",
        "marketing",
        "fiction",
        "roleplay",
        "studying",
        "research",
        "reasoning",
        "math",
        "legal",
        "medical",
        "image-understanding",
        "charts",
        "pdfs",
        "spreadsheets",
        "data-analysis",
        "extraction",
        "presentations",
        "coding",
        "debugging",
        "frontend",
        "local-code",
        "agents",
        "rag",
        "support",
        "translation",
        "language-learning",
        "voice",
        "transcription",
        "images",
        "video",
        "general"
      ],
      "description": "Optional exact task; overrides the broad use case."
    },
    "audience": {
      "type": "string",
      "enum": [
        "everyday",
        "builders"
      ],
      "description": "Assistant-plan guidance or API/local-model guidance."
    },
    "budget": {
      "type": "string",
      "enum": [
        "free",
        "low",
        "mid",
        "unlimited"
      ],
      "description": "For builders, low means at most $5 and mid at most $20 for 1M input + 200k output tokens. free requests open-weight/local candidates, not free hosting."
    },
    "context": {
      "type": "string",
      "enum": [
        "standard",
        "large",
        "huge"
      ],
      "description": "Context requirement. large means 200K+ tokens; huge means 1M+ tokens."
    },
    "openWeight": {
      "type": "string",
      "enum": [
        "yes",
        "no",
        "either"
      ],
      "description": "Whether the user needs an open-weight model."
    },
    "speed": {
      "type": "string",
      "enum": [
        "critical",
        "balanced",
        "quality"
      ],
      "description": "Preference for speed versus best quality."
    },
    "limit": {
      "type": "number",
      "description": "Maximum number of recommendations to return. Defaults to 5."
    }
  },
  "required": [
    "useCase"
  ]
}