Llm Advisor

Gives Claude real-time access to pricing and benchmark data for 336+ LLM and VLM models from five sources including SWE-bench, LM Arena Elo, and OpenCompass. Exposes four tools: get_model_info for detailed specs with code examples, list_top_models for category rankings (coding, math, vision, reasoning), compare_models for side-by-side tables, and recommend_model for filtered suggestions based on price, context window, and capabilities. Data refreshes hourly from public APIs with no authentication needed. Returns compact Markdown tables instead of raw JSON to save context window space. Useful when you need current model performance data that post-dates your assistant's training cutoff, or when comparing options for a specific task with budget constraints.