* GPT-5, Grok-4, Gemini, Kimi-k2 model support, /clear confirmation, /undo output condensed
* Custom voices, Sora API, GPT-5.4, computer tool, websockets, batch images, skills
* Improved model ordering, cloud models no-pull, Claude compaction, thinking levels
* MCP elicitation, /context suggestions, model overrides, /loop command, /claude-api skill
* Context drought, LLM aspirations, hybrid search, Replit Agent 4, AMI Labs, NVIDIA AI Engineers
* MCP conversion helpers, automatic cache, Sonnet 4.6, Opus 4.6 fast/adaptive, Structured Outputs, binary streaming
* NVIDIA NeMo Retriever, Storage Buckets, RL Libraries, Granite 4.0 Speech, Ulysses SP, LeRobot
* GPT 5.4, Gemini 3.1 Flash Lite, Anthropic risk, QuitGPT
* Theme screenshots, return key to enter, chat resume footer, SVG styles, A2A timeout, thinking UI overhaul
---
* Claude Code: Removed Opus 4 and 4.1 from first-party API