> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://docs.caesar.xyz/llms.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.caesar.xyz/_mcp/server.

# Model Selection

The `model` parameter on `POST /research` lets you choose which LLM synthesizes the final answer. If you omit it, Caesar automatically selects a model based on your query.

> **Tip**
>
> Start by omitting `model` to let Caesar auto-select. Set `model` when you need consistent output, a specific latency profile, or a fixed provider.

## Supported models

| Model              | Provider  | Summary                                                                 |
| ------------------ | --------- | ----------------------------------------------------------------------- |
| `gpt-5.2`          | OpenAI    | Highest quality synthesis for complex, multi-step research tasks        |
| `gemini-3.8-flash` | Google    | Default model for deep research, served through Vertex AI               |
| `gemini-3.1-pro`   | Google    | Backward-compatible alias for `gemini-3.8-flash`                        |
| `gemini-3-pro`     | Google    | Advanced reasoning with strong long context performance                 |
| `gemini-3-flash`   | Google    | Best performance for high-volume, low-latency research                  |
| `claude-opus-4.6`  | Anthropic | Strongest Claude-tier synthesis quality for long-form, nuanced analysis |

> **Note**
>
> Prefer `gemini-3.8-flash` for new integrations. Existing requests using `gemini-3.1-pro` remain accepted and now route to `gemini-3.8-flash` through Vertex AI. `gemini-3-pro` continues to select its existing model.

> **Note**
>
> The `model` parameter controls research synthesis. Retrieval and source gathering remain the same.

## When to use each model

#### gpt-5.2

**Best for**

* Complex, multi-step reasoning
* Cross-domain technical synthesis
* High-stakes decisions that need the strongest accuracy

**Trade-offs**

* Typically higher latency than Flash-tier models

#### gemini-3.8-flash

**Best for**

* Deep analysis in code, math, or STEM topics
* Long-context synthesis over large documents or datasets
* Deep research using Caesar's default Gemini model

**Trade-offs**

* Requests using the legacy `gemini-3.1-pro` name use this same model

#### gemini-3-flash

**Best for**

* Large scale processing and batch research
* Low latency, high volume workloads
* Agentic or iterative tasks that need fast turns

**Trade-offs**

* Less depth than Pro or Opus on very complex analysis

#### claude-opus-4.6

**Best for**

* Maximum Claude-tier synthesis quality
* Highly nuanced argumentation and editorial polish
* Long-form outputs where tone and structure matter

**Trade-offs**

* Higher latency than Flash-tier and many non-Opus models

## Example

```json
{
  "query": "Compare major approaches to carbon capture and their performance",
  "model": "gemini-3-flash",
  "reasoning_loops": 2
}
```

## Quick picker

| Goal                                 | Recommended model  |
| ------------------------------------ | ------------------ |
| Fast, high volume research           | `gemini-3-flash`   |
| Deep research with Gemini            | `gemini-3.8-flash` |
| Maximum accuracy on complex research | `gpt-5.2`          |
| Maximum Claude-tier writing quality  | `claude-opus-4.6`  |

## Learn more

* [https://openai.com/index/introducing-gpt-5-2/](https://openai.com/index/introducing-gpt-5-2/)
* [https://ai.google.dev/gemini-api/docs/models](https://ai.google.dev/gemini-api/docs/models)
* [https://www.anthropic.com/news/claude-3-family](https://www.anthropic.com/news/claude-3-family)