Faster Response Problem #142

dd2480467-source · 2026-04-06T04:24:40Z

dd2480467-source
Apr 6, 2026

How can i make it so it responds faster? It takes around 15 seconds to respond i am wondering if i can make it more automatic.

Answered by isair

Hey! Just released a new version which includes performance improvements that should help with this.

Here's what's new:

Thinking mode toggles — You can now disable thinking/reasoning for your LLM in the settings, which can significantly speed up responses.
Faster Whisper transcription — CPU transcription has been optimised, so voice input processing is quicker.

Beyond that, response speed mainly depends on which LLM model you're running and your hardware. A few tips:

Use a smaller model — If you're running a large model (e.g. 14B+ parameters), try a smaller one. Something in the 3B-7B range will respond much faster.
Disable thinking mode — In settings, turn off thinking/extended thinki…

isair · 2026-04-08T20:15:23Z

Hey! Just released a new version which includes performance improvements that should help with this.

Here's what's new:

Thinking mode toggles — You can now disable thinking/reasoning for your LLM in the settings, which can significantly speed up responses.
Faster Whisper transcription — CPU transcription has been optimised, so voice input processing is quicker.

Beyond that, response speed mainly depends on which LLM model you're running and your hardware. A few tips:

Use a smaller model — If you're running a large model (e.g. 14B+ parameters), try a smaller one. Something in the 3B-7B range will respond much faster.
Disable thinking mode — In settings, turn off thinking/extended thinking for your model. This skips the internal reasoning step and gives you a faster reply.
GPU acceleration — If you have a GPU, make sure Ollama is using it. This makes the biggest difference.

Please update and let me know if things improve! 🚀

0 replies