- Published on
Notion Lets You Choose Your AI Model: Workers, Multi-Model, and Audio Notes Explained
If you use Notion, you have probably tried Notion AI. If it felt "good enough, nothing special," the August update is worth another look.
This month, Notion added three features. Each is meaningful on its own, but together they point in a clear direction. Notion is transitioning from a note-taking app to an AI workflow platform.
Table of Contents
- Choose Your AI Model β GPT-5.2 vs Claude Opus vs Gemini 3
- Notion Workers β External API Integration Without Custom Code
- Audio β Meeting Notes Automation β Upload and Done
- Usage Tracking and Agent Sharing β Quietly Important
- Which Model Should You Use β A Practical Selection Guide
1. Choose Your AI Model β GPT-5.2 vs Claude Opus vs Gemini 3
You can now choose which AI brain powers your Notion conversations.
The most talked-about change in Notion's August update is the AI model selection feature. When starting a conversation, a dropdown lets you choose your model.
Currently supported models:
- GPT-5.2 (OpenAI) β Strong for general writing and fast drafts
- Claude Opus 4.5 (Anthropic) β Strong for long document analysis and complex reasoning
- Gemini 3 (Google) β Strong for multimodal tasks and Google Workspace integration
The model selection UI is designed intuitively. A "For your hardest tasks" section groups the most capable models together, and frequently used models can be pinned as favorites at the top.
Why does this matter?
Until now, AI tools were black boxes using an internal model. Notion introducing multi-model selection signals that the choice of AI quality is moving to the user. Use Claude for complex strategy documents, GPT for quick drafts, Gemini for data analysis β task-level optimization becomes possible.
2. Notion Workers β External API Integration Without Custom Code
Notion can now connect directly to external services.
Officially announced in August, Notion Workers lets you deploy code that calls external APIs and processes data from within a Notion page.
Here is the simple version:
- Before: Getting external data into Notion required tools like Zapier or Make
- After Notion Workers: Deploy code via Notion's own CLI and call APIs directly
Real usage examples:
- When a specific keyword appears in Slack, automatically add an item to a Notion database
- Periodically pull Google Sheets data into Notion
- Sync customer CRM data with Notion project pages in real time
Technically, code is deployed via CLI and runs in a sandboxed environment. Windows (x64) support has also been added.
Pricing: Free through August 11, 2026, then metered billing via Notion credits.
| Feature | Before Notion Workers | After |
|---|---|---|
| External API integration | Separate automation tool required | Deploy directly inside Notion |
| Maintenance | Manage multiple services | Consolidated into Notion |
| Learning curve | Need to learn automation tools | Works with CLI knowledge |
3. Audio β Meeting Notes Automation β Upload and Done
If you recorded the meeting, the AI writes the notes.
A quiet but highly practical feature arrived in this update. Upload an audio file to Notion meeting notes and AI automatically:
- Generates a full transcript
- Summarizes key discussion points
- Organizes action items
Upload the recording after the meeting. AI handles the rest.
Education applications:
- Record a class β auto-summarize lesson content
- Record parent meetings β automatically log key points
- Record staff meetings β auto-organize decisions and role assignments
From an edtech perspective, this is a small but meaningful step toward freeing teachers from administrative burden. The 30 minutes spent writing meeting notes can go to student feedback instead.
4. Usage Tracking and Agent Sharing β Quietly Important
Features that are easy to overlook but matter significantly for team use.
AI Usage Tracking
Settings β Notion AI β Usage now shows your AI consumption. It tracks both a rolling six-hour window and your monthly billing cycle. For teams on paid plans, this directly supports budget management.
Pass Documents Directly to Agents
Since August 7, Notion pages and databases can be shared directly with an Agent from the Share menu. Previously, referencing a document in an agent conversation required copy-pasting the content.
On August 11, embedding agent conversations inside pages was also added. You can now run a project-specific AI agent directly from a project page and log results in the same page.
5. Which Model Should You Use β A Practical Selection Guide
Matching the model to the task type maximizes value for cost.
| Task Type | Recommended Model | Reason |
|---|---|---|
| Quick drafts, simple summaries | GPT-5.2 | Speed and natural writing style |
| Complex contracts, long report reviews | Claude Opus 4.5 | Long context handling, fine-grained reasoning |
| Data-heavy tasks, Google-connected work | Gemini 3 | Multimodal, Workspace integration |
| Difficult strategic planning | Claude Opus 4.5 or top GPT model | Use the "For your hardest tasks" section |
Practical tip β use Favorites
Pin two frequently used models as Favorites so you can select them from the top without opening the dropdown every time. Consider setting a "daily work default" and an "analysis" model as your two pinned favorites.
Closing
The thread connecting all three features is integration.
- Multi-model selection β the best AI, inside Notion
- Notion Workers β external services pulled into Notion
- Audio β transcript β action items, all completed inside Notion
Notion's direction is clear: become the hub for all work. A transition from knowledge management tool to AI operations platform.
Which AI model do you think you will reach for most often in Notion? Let me know in the comments!
Related Posts
- NotebookLM Is Gone? 5 Key Changes in the Gemini Notebook Rebrand
- Vibe Coding 2026: 63% of People Building Software Have No Coding Background
Source: