Method, data quality, and limits
The race view compares public capability measurements for major Western and Chinese AI labs. It does not claim that a benchmark lead determines who will achieve AGI first.
Epoch’s Capabilities Index combines scores from more than 50 benchmarks into a general capability scale. We use the highest dated score in each ecosystem as its composite frontier.
ECI is not a percentage, AGI probability, safety score, or arrival forecast.
Models are grouped by the country and primary lab reported in the source. Western includes US and European labs in the tracked roster; China includes mainland Chinese labs.
Country and organization fields come from the public dataset. Cross-border partnerships and multinational teams make any geographic grouping imperfect, so this is an ecosystem lens—not a claim about individual researchers.
The model roster keeps the three highest-scoring named models per tracked lab and adds the lab’s strongest open-weight model when it is not already represented.
Reasoning-effort variants sharing the same source Model name are deduplicated. The highest-scoring representative is retained, with explicit high/max variants preferred on tied scores.
We use “open weights,” not “open source.” A model qualifies when the dataset’s accessibility field begins with Open weights. Restrictions, non-commercial terms, and license obligations are preserved in the model table.
Rows without valid score, date, or ecosystem identity are excluded. A missing or non-positive ecosystem result is displayed as “no comparable result,” never silently converted to a zero.
124 ECI rows are outside the scored race calculation because they lack a valid score or tracked ecosystem classification.
The workflow checks Epoch’s public archive daily at 04:17 UTC, rebuilds both evidence snapshots, and deploys the static Cloudflare Pages output when the production workflow is connected.
Current snapshot: 2026-07-25. Latest release date observed in the ECI source: 2026-07-24.
Quality notes
Primary references