sota-tracker-mcp
romancircus 作
Browser & Automation概要
**The definitive open-source database of State-of-the-Art AI models.**.
説明
このスキルでできること
A daily-updated database of state-of-the-art AI models across categories including LLMs, image generation, video, and speech. Data is pulled from LMArena Elo rankings, Artificial Analysis benchmarks, and HuggingFace. Accessible via JSON/CSV downloads, SQLite queries, a REST API, or an optional MCP server.
Aggregates three major leaderboards into one queryable database updated daily, replacing the need to manually check LMArena, Artificial Analysis, and HuggingFace separately.
使いどころ
- Checking which LLM ranks highest before choosing a model for a new project
- Querying a list of deprecated models to avoid in agent configurations
- Comparing two models side by side using current benchmark data
- Auto-updating a CLAUDE.md file with current top models each morning
- Getting a task-specific model recommendation for coding, chat, or image generation
ワークフロー例
User asks: What are the top local LLMs for coding right now?
- 1
Query the REST API endpoint GET /api/v1/models?category=llm_coding
- 2
Parse the returned JSON for models sorted by sota_rank
- 3
Check the forbidden list to exclude deprecated or outdated models
- 4
Return the ranked model names with their benchmark context
Ordered list of current top coding LLMs such as Qwen3-Coder and DeepSeek-V3 with ranking details
インストール
clawhub install romancircus/sota-tracker-mcpCLIコマンドを実行するか、Claude Code、Cursor、Codexなどお使いのコーディングエージェントにプロンプトを貼り付けてください。