Generation comparisons

Alignment Index

GPT series
707580859084.2Sol-5.687.6Sol-687.9Sol-6.1
Claude Opus series
707580859074.5Opus 4.875.6Opus 583.2Opus 5.5
Gemini Flash series
707580859074.03.6 Flash74.43.7 Flash75.83.8 Flash
Grok series
707580859079.1Grok 4.580.8Grok 4.682.7Grok 4.7
Figure 7. Alignment Index and signal detection rates across successive model generations. Each panel compares three generations within a model family. Higher index scores and lower detection rates indicate greater alignment. Error bars show 95% confidence intervals. Use the tabs to switch between the index and individual signals.