0% idle CPU usage has been achieved internally.

Fix cpu usage.

• Optimize idle cpu usage
• Fix temperature setting for Kimi K2.5

Fix reasoning UI for OpenAI, #1207

Tweaks.

Fix reasoning UI for OpenRouter gemini-3-flash

Add gpt-image-1.5 in OpenAI and AiHubMix

Fix GPT 5.2 xhigh reasoning

fix #1201

Fix command center

Fix chat restoring, #1197

fix invalid peer certificate error

fix http: use rustls

Fix styling and description for xhigh reasoning effort

• Support new GPT-5.2 reasoning effort: xhigh
• Remember last opened chat and restore on app startup

Fix web search.

Update default models for web search / chat title

Fix web search #1172

• Hide image generated in thought process, #1170
• Update MCP SDK to the latest version, #1169
• Use Command+Number (mac) or Ctrl+Number (others) to switch assistant

Fix Anthropic / Claude 4.5 Opus

Fix duplicated image output in Gemini 3 Pro Image

Fix image width.

Fix native web search with Gemini 3

• Support GPT-5.1 and GPT-5.1 Codex
• Internal: Use file system instead of IndexedDB for caching

minor mcp improvements

• Allow to use custom SVG as assistant icon (preview)
• Add new provider: Baseten

Fix ListenHub: include chinese voices

• New Text-to-speech provider: ListenHub
• No longer shows the "Favorites" model group when the "All" tab is selected in the model dropdown

Fix assistant sorting on Windows, #1146

Add Web Fetch extension: Settings > Extensions > Web Fetch, you can now set a Jina AI API key for higher rate limits