Change Log

In Deep Think mode, automatically detects ambiguous or incomplete questions and asks follow-up questions with option chips to help you describe your needs precisely, up to 5 rounds.

During multi-model debate, scoring begins at most 30 seconds after the first model completes, significantly reducing wait time.

Export AI responses as PDF, DOCX, or TXT for archiving and sharing.

Customize LLM model selection and AI response language. Settings sync across devices.

Reduces AI hallucination through native Web Search, fact-checking, and source citation mechanisms with reference links.

Backend automatically cancels incomplete LLM tasks and frontend cleans up residual streaming messages to prevent garbled output when switching conversations.

Upgraded to the latest Claude 4.6 Sonnet model with a new Skill application tab showing applied analysis frameworks.

Define custom enterprise terminology so AI responses will correctly use your specialized terms, improving response accuracy.

Supports large text input limits and a context sliding window mechanism for effective long conversation management.