FinRetrieval (MCP Tools): leaderboard
FinRetrieval (Daloopa, 2026) asks agents for 500 specific numeric values from company filings, earnings releases and investor presentations; an LLM judge marks each answer correct or incorrect against the ground truth. Each row is a vendor agent stack (model plus its official agent SDK) with web search and a structured financial-data API over MCP, in base or extended-reasoning mode; accuracy (%), higher is better.
Metric: Accuracy (%). Source: huggingface.co. Saturation forecast: Around February 2027. 8 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Pro (Preview) (High) | 90.8 | #111 (Gemini 3 Pro (Preview)) |
| 2 | GPT-5.2 (High) | 89.4 | #172 (GPT-5.2) |
| 3 | Gemini 3 Pro (Preview) (Low) | 87.6 | #111 (Gemini 3 Pro (Preview)) |
| 4 | GPT-5.2 | 80.4 | #172 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=finretrieval-mcp-tools · How It Works · Data refreshed daily, snapshot 2026-11-12.