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

# DeepSeek

DeepSeek's V4 models offer strong reasoning at substantially lower cost than equivalent OpenAI or Anthropic tiers. The wire is OpenAI-compatible (chat completions, streaming, tools), so the integration is straightforward.

## Get credentials

1. Sign in at [https://platform.deepseek.com](https://platform.deepseek.com).
2. Open **API keys** in the dashboard (direct link: [https://platform.deepseek.com/api\_keys](https://platform.deepseek.com/api_keys)).
3. Click **Create new API key**, copy the value.

## Configure in jambonz

In the portal: **Account → LLM Services → + Add LLM Service → DeepSeek**.

**`API Key`** `string` — required

The key from DeepSeek's dashboard.

---

**`Base URL`** `string`

Defaults to `https://api.deepseek.com/v1`. Override for self-hosted DeepSeek deployments or proxy endpoints.

---

Click **Test** to verify.

## Use in an agent verb

```js
session.agent({
  llm: {
    vendor: 'deepseek',
    model: 'deepseek-v4-flash',
    llmOptions: {
      systemPrompt: 'You are a helpful voice assistant.',
    },
  },
  stt: { vendor: 'deepgram', language: 'en-US' },
  tts: { vendor: 'cartesia', voice: 'sonic-english' },
  turnDetection: 'krisp',
  bargeIn: { enable: true },
  actionHook: '/agent-complete',
}).send();
```

## Available Models

See DeepSeek's [models & pricing page](https://api-docs.deepseek.com/quick_start/pricing) for the full list and current pricing. Common picks:

| Model               | Notes                               |
| ------------------- | ----------------------------------- |
| `deepseek-v4-flash` | Default — fast, cheap, tool-capable |
| `deepseek-v4-pro`   | Higher quality at higher cost       |

## Quirks & errors

> **Note**
>
> **`deepseek-chat` and `deepseek-reasoner` are deprecated** — older V3-era ids that the manifest flags as `deprecated: true`. They still work in some accounts but are scheduled for removal. Switch to `deepseek-v4-flash` or `deepseek-v4-pro`.

> **Warning**
>
> **Tool-calling reliability is weaker than OpenAI/Anthropic** on shorter prompts. If your agent depends on consistent tool invocation, test thoroughly with realistic prompts. For complex tool flows, consider Claude or GPT-4o.

> **Warning**
>
> **`401`** — key revoked or copied with whitespace. Regenerate at [platform.deepseek.com/api\_keys](https://platform.deepseek.com/api_keys).