Methodology
How to read the AIUpdateWatch dashboards
Every dashboard separates official data, public API data, benchmark signals, estimates, news signals, and manual observations so readers know how much weight to give each row.
Data quality rules
- Official provider pages carry more weight than social posts, community comments, or secondary summaries.
- Benchmark rows should be treated as signals, not universal proof that one model is best for every use case.
- Pricing rows should be checked against the provider’s latest pricing page before making a large buying decision.
- Estimator dashboards must be read as assumptions-based comparisons, not confirmed provider cost or profit data.
- Manual GEO, citation, and drift observations are useful samples, not complete market coverage.
- News and community mentions show attention, not guaranteed technical capability.
Benchmark and value signals
Frontier Leaderboard, Open-Weight Leaderboard, and Intelligence Per Dollar compare public quality signals with cost and capability data. Use them to build a shortlist, then check the exact model against your own task before committing.
Research, launch, and news signals
Research dashboards show what is being published, GitHub dashboards show what developers are paying attention to, and launch/news dashboards show what is gaining public attention. A strong signal across all three is more meaningful than a spike in only one place.
Cost and infrastructure estimates
Local vs Cloud Cost, Token Margin Tracker, and GPU Availability are designed for comparison. They help readers ask better questions about hardware, API usage, rented GPUs, power costs, and scale, but they should not be read as guaranteed quotes.
Visibility and model-behavior observations
Share of Model, Citation Source Tracker, AI Sentiment Map, and Model Drift Monitor are observation dashboards. They are useful for repeatable checks: whether a brand appears, which sources are cited, how a topic is described, and whether repeated model answers change over time.