Skip to content

v0.9.1 - thinking on by default (chat), all families

Choose a tag to compare

@srvsngh99 srvsngh99 released this 17 Jun 14:05
· 185 commits to main since this release

KrillLM v0.9.1 turns the reasoning ("thinking") channel on by default in interactive chat, with a toggle and a config default, and makes it work both ways across every thinking-capable model we ship.

  • On by default in krillm run (TUI + --classic REPL) for models with a thinking channel. Toggle per session with /think on|off or Ctrl-T; a think dot shows in the footer. Default set by thinking = true|false in ~/.krillm/config.toml (or KRILL_ENABLE_THINKING).
  • Both-ways toggle, verified across families: Qwen 3 (enable_thinking, defaults on - now genuinely turns off too), the Gemma-4 channel templates (stock Gemma-4 + the coder fine-tune), and DeepSeek-R1 distills (always reason - no off switch, so the toggle is a no-op there). Non-reasoning models (Llama, Mistral, Phi, Qwen2.5, Gemma 2) are unaffected.
  • Server/API and single-shot are unchanged - they make no explicit decision, so the model's template default is preserved (no silent API change).

Install: brew tap srvsngh99/krillm && brew install krillm (or brew upgrade krillm)
Asset: krillm-0.9.1-arm64-apple-macos.tar.gz (Apple Silicon). Changelog: [0.9.1].