gpt-6IMAGE
Letting GPT-6 Astra and Opus 5.5 write code to play StarCraft
A tester let two models write code and then play StarCraft against each other, finding them close but with fast capability decay, suggesting gaps larger than the Intelligence Index shows.
kai
@kaimcpheetersArchived 2026.09.27Originally posted 2026.09.27

kai @kaimcpheetersOriginally posted 2026.09.27
GPT-6 Astra and Claude Opus 5.5 can be competitive in StarCraft once they are allowed to write code. However, the field falls off fast - suggesting bigger differences in capabilities than other benchmarks like @ArtificialAnlys Intelligence Index might suggest
x.com/kaimcpheeters/status/2103880194930966869Archived 2026.09.27 10:07