Open data
Query every poll we've ever run
Don't take our rankings on trust — audit them. Every answer ChatGPT, Claude, Gemini and Grok have given us is on the record: timestamped, failures included, nothing overwritten. It's served through a free JSON API — no key, no auth, CORS open, CC BY 4.0 with attribution to modelsagree.com.
1,425
categories
4
models polled
weekly
re-poll cadence
Endpoints
| /api/v1/search?q=llm+observability | find a category (or brand) by name |
| /api/v1/best/{slug}.json | full 4-model verdict for one category |
| /api/v1/history?slug={slug}&limit=100 | the raw poll record: every answer over time, failures included |
| /api/v1/categories.json | every category with its API url |
| /api/v1/mentions.json | all-time mindshare leaderboard |
Start with the agent guide — usage recipes with real response examples — or the OpenAPI spec.
Read this before citing it
These are answers from the consumer surfaces of each model, one fresh session per question, at whatever settings each product ships. A single answer is noise; the repeated weekly record is the signal — which is why the history endpoint exists. This data measures what AI assistants recommend, not independent product testing. Full protocol on the methodology page. Last refresh: August 10, 2026.