Improvement
Patch notes · August 13
The complete notes for this update.
Improvements
- Hey all! We have two model updates to share with you. **DeepSeek V4 Flash 0731** The experimental DeepSeek model, DeepSeek V4 Flash 0731, performed very well in our model testing, so we’re making it a permanent model, as well as the new default for Flash! You can still find and use the older Flash version ("1.0.0") in the dropdown Version Selector. **GLM 4.6 Context Correction** We’ll also be correcting an issue with GLM 4.6, previously known as Raven. GLM 4.6 and 4.7 are meant to offer twice the context of 5.1 and 5.2. However, a bug caused that doubling to be applied twice to GLM 4.6, giving it twice as much context as intended—on top of its already doubled amount. We’re now adjusting it to match GLM 4.7. This also means that GLM 4.6 will be **available to Journey players using credits**, and **Context Optimization will be able to be turned off**. GLM 4.6 will still be deprecated; this change simply makes it consistent with GLM 4.7 while it remains available. Please let us know if anything looks off after these updates. Thank you!