Patch Notes B0.042 - GLM 5.3 Flash Back in Core
- Sorry for the whiplash! We removed the model from core due to complaints about how slow it was and increased timeout rates. After hearing feedback from users in Discord, we've decided add GLM 5.3 Flash back to core. Users using GLM 5.3 flash should understand that it's just a slow model - even the fasted providers still end up being fairly slow sometimes.
- Added a Slow tag and a confirmation before choosing slow models, explaining that it is best suited to times when accuracy matters more than speed.
- Prevented models marked Slow from being set as a world's recommended model, while keeping them available for players to choose themselves.
- You should see improvements in Deepseek v4 Flash as well, we have tuned the providers once again.