AGI
SpaceX’s Grok against the best models from every other lab, over time, on the benchmarks closest to general intelligence. The frontier is the best score to date; the gap is how far behind it Grok has been.
Loading
…
The frontier, and where Grok sits
Every frontier model at release. The line is the best score to date; Grok’s best to date is in its colour.
Grok against the frontier, over time
Grok’s best score to date as a share of the frontier’s, on each benchmark.
The compute behind them
Training compute of notable models (Epoch AI estimates), and xAI’s Colossus.
His AGI predictions
Dated, scored as on the record.
…