Aider v0.86.0 ships model support for four new LLMs at once — GPT-5, Grok-4, Gemini 2.5 Flash Lite, and Kimi K2. The open-source AI coding assistant also wrote 88% of the code in this release.
New model additions:
- All GPT-5 models — full support, no qualifier
- Grok-4 via
xai/grok-4andopenrouter/x-ai/grok-4 - Gemini 2.5 Flash Lite preview (
gemini/gemini-2.5-flash-lite-preview-06-17) - Kimi K2 via
openrouter/moonshotai/kimi-k2
The model wave matters because Aider is one of the fastest open-source tools to integrate new providers. Users who want to benchmark GPT-5 or Grok-4 against Claude on real coding tasks can now do it inside the same terminal workflow, switching with a single /model flag.
Beyond model support, the release cleans up several rough edges. /clear now prints confirmation text so you’re not guessing whether chat history actually reset. /undo trims commit-message output to the first line only — a small fix that makes multi-commit sessions far easier to scan. A bug where new model settings silently failed to override existing ones is fixed. Dependencies bump to litellm==1.75.0.
NOTE
The no-arg /model command now displays model announcements inline — useful for catching provider-specific notes without leaving your session.
88% AI-authored code in a single release is the number worth watching. Aider has been tracking this metric publicly for months, and it keeps climbing. The tool is, increasingly, building itself.