Pick a model
Choose among 30+ models from a dozen providers, switch mid-thread, and fall back automatically on provider errors.
docAnalyzer doesn't lock you into one model. The catalog includes 30+ models from a dozen providers (Anthropic, OpenAI, Google, DeepSeek, Mistral, Qwen, Groq, and more) all behind one normalized interface. Pick a model in chat settings; switch when you want; if a provider has a bad day, docAnalyzer falls back to the next model in your chain.
The baseline set vs. Lab models
The catalog splits into two tiers:
- Baseline: available on every plan. The mainstream catalog — everyday Q&A, summarization, long-context analysis, reasoning, and structured-output tasks, picked for a good speed/quality/cost balance.
- Lab models: an optional toggle in your account settings (paid plans) that adds models from smaller and newer providers, for experimenting beyond the standard set.
For most questions, the baseline is plenty. Reach for a Lab model when you want to try a new provider or compare how different model families handle the same task. The fine-grained knobs (thinking effort, context density, answer length) are separate from the catalog — they live in chat settings under Advanced (editable on paid plans).
Switch mid-thread
You don't have to commit to a model at the start of a chat. Open the model picker in any Focus or Co-work and switch. The conversation continues, tool state is preserved, citation tokens stay valid. This is useful when:
- A cheaper model handles the easy part of a thread and you want to escalate for the hard turn.
- You want to compare two models' responses to the same question without losing the lead-up.
- A provider is having an outage and you want to retry the same question elsewhere.
Automatic fallback on rate-limit errors
If the model you picked hits a rate limit mid-turn, docAnalyzer rotates to a compatible fallback model instead of failing the conversation. You'll see a small notice that the model changed; the thread continues.
Which model for which task
Rough guidance, since your mileage will vary by document type:
| Task | Lean toward |
|---|---|
| Quick Q&A, summaries | Fast, mainstream model in baseline |
| Long-document analysis (50+ pages) | Long-context model |
| Multi-step reasoning, math | A reasoning model; thinking effort dial helps |
| Code or structured output | A code-capable model |
| Multilingual content | Test a couple; coverage varies |
| Visual content (charts, figures) | Multimodal-capable model (required for View Images) |
If you're not sure, just try one; switching is cheap.
Costs
Most baseline models run on docAnalyzer's account and are included in your plan's prompt fair-use limits (see your plan). Premium models meter against credits; usage shows up in your credit dashboard.
What's next
- Set persistent instructions: tell every chat how you like to work.
- Tune source adherence: stricter grounding for high-stakes answers.
- How credits work: what premium-model usage costs.