Strata fixes
- Fixed a Windows startup error that prevented DeepSeek V4.1 Flash from opening GGUF shards that were already on disk. Existing downloads can be reused.
- Adjust the context window directly in chat for Strata, FastFlowLM, GenieX, LiteRT-LM, vLLM, SGLang, and TensorRT-LLM. Where the engine reports the model's maximum, the slider respects it. The chat shows when the active model is stopping and reloading.
- The Model Hub now groups all three quantizations of DeepSeek-V4.1-Flash-1017 Uncensored in one Strata card and labels each download option.
- Includes Atomic Chat Core v0.24.7.
Other fixes
- Factory reset now clears generated videos and engine data.
Contributors
Thanks to @Ooooze.