Oct 28, 2026 کو اپ ڈیٹ کیا گیا
- First words come sooner on new chats and after app open
- Qwen3.5 chats keep their memory between turns; long chats no longer re-read everything
- Thinking models answer after their thinking budget, on MNN and GGUF
- Brazilian Portuguese, starter characters in your language, translation fixes
- Fixes: load crashes, damaged model files refused, web search on DeepSeek, Mistral, Gemma 4 and Llama, Top-K 0 crash, long paste freeze
