Model Catalog and Plan Availability
Whisperer uses different AI models for different tasks. Which models are available depends on your plan. This page explains what's what.
Roles and model families
Each type of task is served by a separate model "role." For each role you can choose a specific model in settings.
| Role | Task |
|---|---|
| responses | Regular answers to your questions in real time |
| coding | Writing and explaining code |
| vision | Analysing screenshots and images |
| system_design | Structured responses with diagrams |
| generation | Building the meeting mind-map and analytics |
| transcription | Speech recognition (Whisper) |
| embedding | Knowledge-base search (RAG) |
The transcription and embedding roles are fixed and cannot be changed; responses, coding, vision, and system_design are configurable.
Available model families
Whisperer provides models from the following families:
- Claude (Anthropic) — Haiku, Sonnet, Opus
- GPT (OpenAI) — from mini to full-size
- DeepSeek — chat and coder variants
- Qwen — chat and coder
- Gemini (Google)
- Grok (xAI)
- Llama (Meta, via provider)
Availability by plan
| Model group | Free | Start | Pro | Max |
|---|---|---|---|---|
| Basic models | ✓ | ✓ | ✓ | ✓ |
| Extended set | — | ✓ | ✓ | ✓ |
| Frontier models | — | — | — | ✓ |
Frontier models are the flagship variants of each family (e.g., Claude Opus, top-tier GPT). They deliver the highest response quality but are available only on the Max plan.
📸 [Screenshot: model selection screen in settings — list with plan availability labels]
How to choose a model
- Open the web dashboard → Settings → Models.
- Choose a model for each role from the available list.
- Changes apply to the next session.
Models unavailable on your plan appear greyed out — click them to see which plan unlocks them.
Model selection tips
- For quick hints: mid-tier models within a family — balance of speed and quality.
- For system design and complex tasks: if depth matters, upgrade to Max for frontier access.
- For code: coding-optimised models (DeepSeek Coder, GPT + code-oriented Claude) produce more accurate code.
- For multi-screenshot: make sure the vision model you select supports multi-image input; check this in settings.
Common issues
| Issue | Cause | Fix |
|---|---|---|
| Desired model is greyed out | Not available on current plan | Upgrade your plan or pick an available alternative |
| Model changed after upgrading | Settings cache | Refresh the page and reopen settings |
| Response quality below expectations | Basic model selected | Try a higher-tier model in the same family |