Models & Settings
Nebula Writer lets you choose the model and reasoning level for each assistant session while keeping a sticky default for future work.
Free accounts include 40 agent requests per month and are locked to GPT-5.5. Pro accounts can choose from all enabled managed models and use Pro-only search and cloud collaboration features.
When to Use This Page
- You want to understand the model picker in the assistant
- You switch between fast editing, deep research, and complex project work
- You want defaults that stay consistent across sessions
Model Picker
Open the assistant panel and use the model menu in the message bar.
The current model menu includes:
| Model | Good For |
|---|---|
| GPT-5.5 | Default high-capability writing, research, and editing |
| GPT-5.4 | Strong everyday drafting and code-aware work |
| GPT-5.2 | Reliable general writing and review |
| Claude Sonnet 4.6 | Long-form prose, critique, and careful revision |
| Claude Opus 4.6 | Deep reasoning and complex document restructuring |
| Gemini 3.1 Pro | Broad research and multimodal reasoning tasks |
| Gemini 3 Pro | General advanced research and editing |
| Gemini 3 Flash | Faster short tasks |
| Kimi K2.5 | Alternative long-context reasoning |
| GLM-5 | Alternative general assistant work |
The model you select becomes the assistant's default model. Free accounts will stay on GPT-5.5; Pro accounts can choose another enabled model before sending the prompt, and Nebula Writer will keep that default until you change it again.
Reasoning Level
The reasoning menu controls how much effort the assistant should spend before answering.
| Level | Use It For |
|---|---|
| None | Simple transformations and quick factual questions |
| Low | Everyday edits, summaries, and short planning |
| Medium | Multi-file comparisons and careful revisions |
| High | Research synthesis, structural edits, and higher-risk changes |
Higher reasoning can produce better plans for complex work, but it may be slower and may use more quota.
Choosing the Right Model
Use a fast setup for:
- Short rewrites
- Formatting cleanups
- Summaries of the active document
- Simple questions about a file
Use a stronger setup for:
- Comparing multiple files
- Editing DOCX, spreadsheet, or PPTX structure
- Generating Python analysis plans
- Research synthesis with citations
- Major rewrites where factual accuracy matters
Practical Workflows
Editing a Draft
- Choose a strong general model such as GPT-5.5.
- Keep reasoning on Low or Medium.
- Select the paragraph or section you want edited.
- Ask for a focused change.
Rewrite the selected section for clarity. Keep all citations and technical claims.
Research Synthesis
- Choose a higher-capability model.
- Set reasoning to Medium or High.
- Mention the files that matter.
- Ask the assistant to plan before editing.
Read @paper.tex and @research-brief.md. Identify weak claims, suggest sources, and wait before editing.
Data Work
- Open the spreadsheet or Python file.
- Attach or mention the relevant workbook.
- Ask for a short analysis plan first.
Use @research-dashboard.xlsx to identify the strongest trend and propose one chart for the paper.
Account and Quota Notes
AI features require sign-in and internet access. Local editing, project browsing, Markdown, LaTeX source editing, DOCX viewing/editing, spreadsheets, and presentations can still use local files, but assistant calls depend on the signed-in account and available quota.
Free accounts include 40 agent requests per month on GPT-5.5. Pro accounts unlock other enabled managed models, search/research tools, and cloud write/collaboration features. When quota is limited, keep prompts narrow and focused so each agent request does useful work.
Best Practices
- Pick the model before starting a long task
- Use Medium or High reasoning for multi-file or high-stakes edits
- Keep simple copyedits on lower reasoning
- Mention files explicitly with
@ - Review diffs before accepting AI changes
- Change the model when the task changes, not after the answer feels wrong