Patch Notes 26.04 - Grok 4.5, Model Refresh
We’ve refreshed some of the models being used in the backend for newer more intelligent models, a few in particular to note:
Gemini 3.1 Pro has been replaced by Grok 4.5: The Gemini Pro models have been underperforming lately after some changes from Google. Grok 4.5 is more intelligent, better suited for role-play, and slightly cheaper at 5 credits a turn compared to 6.
DeepSeek V4 Flash has been refreshed to the latest V4 Flash-0731. The latest V4 Flash is much smarter than the previous version and is on par with DeepSeek V4 Pro. Give it a try if you typically use Fenix, this may be your new go to standard model.
The models behind turn based combat have been upgraded to newer, smarter models; the combat harness is still the same, but things should run a little bit more smoothly with fewer AI logic errors.