LLM providers and prompts
How to connect is in Connect an LLM. This page collects the differences that help you choose.
Providers
| Provider | Models | Notes |
|---|---|---|
| Amazon Bedrock | Nova Micro, Lite, Pro, Premier; Claude Haiku 4.5, Sonnet 5, Opus 5 | One key for Nova and Claude. Round trips for short input were Haiku 1.9s, Sonnet 2.7s, Opus 3.3s (us-east-1). |
| Anthropic Claude | Four Claude models | Faster than the same models through Bedrock: Haiku 1.3s, Sonnet 1.7s, Opus 2.7s. |
| Google Gemini | Default 3.5 Flash-Lite | The most stable on the free tier. Often leaves out periods. |
| OpenAI (GPT) | GPT models | Uses the Codex CLI sign-in (codex login). |
| Groq Cloud | Groq models | Uses the same key as Groq STT. |
| Local MLX | A downloaded model | Text never leaves your Mac. |
These timings were measured in one development environment and vary with network and region.
Correction modes
| Mode | What it does | Original protection |
|---|---|---|
| Standard (STT Correction) | Fixes misrecognitions, spacing, and English term spelling | If more than 50% of words change, the original is used. |
| Filler Removal | Removes fillers | Same as above. |
| Structured | Organizes into lists and paragraphs | Not applied, since changing the shape is expected. |
| Custom | Uses the prompt template you choose | Not applied. |
After correction, English terms that were in the original but disappeared are restored, and the Dictionary corrections are applied once more.
Prompt templates
Edit templates in Settings → LLM. There are three built-in templates:
| Template | Use |
|---|---|
| General correction | Polishes dictation. |
| LLM prompt | Turns a request you described out loud into a structure that works well when pasted into an LLM. |
| Meeting notes | Used for the final pass of the minutes output. |
- Set a default template for dictation and one for minutes.
- Preview results with sample input.
- Deleting a template asks for confirmation. Deleted built-in templates come back with
Restore built-ins; user templates cannot be recovered. - Updating the app does not change built-in template text you already have. To get the new text, reset the template to its default in the editor.
What is sent to the LLM
Correction sends two things:
- A system prompt: the mode or template text, input-handling instructions, Dictionary corrections, and the active word list
- The text to correct
Sogon doesn't send previous utterances or conversation history, and it doesn't send screen region images to the LLM either.