<!-- Lazu · One API key for GPT, Claude, Gemini and DeepSeek · https://lazu.ai/blog/one-key-every-model -->

# One API key for GPT, Claude, Gemini and DeepSeek

2026-09-30 · Lazu team

Every model vendor you add usually means another SDK, another API key and another invoice. Just comparing GPT, Claude, Gemini and DeepSeek on the same prompt can eat an afternoon of setup.

Lazu is an OpenAI-compatible AI gateway: one key and one base URL reach all of them, and switching models means changing the `model` field. This guide takes five minutes, from creating a key to switching models to reconciling the bill.

## 1. Create a key

Sign in and create a key under [Console › API Keys](https://lazu.ai/console/token). It looks like `sk-lazu-…`. Use one key per app and environment; each can be limited to certain models, a quota, an expiry date and an IP allowlist.

```bash
export LAZU_API_KEY="sk-lazu-..."
```

## 2. See which models you can use

The catalog endpoint lists the models this key can call, the endpoints each supports and the current price:

```bash
curl --fail-with-body https://api.lazu.ai/api/models/catalog \
  -H "Authorization: Bearer $LAZU_API_KEY" \
  | jq '.data[] | {id, supported_endpoints}'
```

You can also browse the [model catalog](/models): every model has its own page with context length, capabilities and prices.

## 3. Call any model with the OpenAI SDK

If your project already uses the OpenAI SDK, point `base_url` at Lazu:

```python
from openai import OpenAI

client = OpenAI(
    api_key="sk-lazu-...",
    base_url="https://api.lazu.ai/v1",
)

for model in ["gpt-5.6-sol", "claude-opus-5-5", "gemini-3.8-flash", "deepseek-v4-pro"]:
    response = client.chat.completions.create(
        model=model,
        messages=[{"role": "user", "content": "Introduce yourself in one sentence."}],
    )
    print(model, "→", response.choices[0].message.content)
```

Same code, same key, four vendors answering in turn. The same works from TypeScript, Go or any OpenAI client that accepts a custom base URL.

## 4. Already on the Anthropic or Gemini SDK?

No need to rewrite to the OpenAI format. Lazu accepts all three authentication styles:

| API style | Header |
| --- | --- |
| OpenAI-compatible | `Authorization: Bearer sk-lazu-...` |
| Anthropic native | `x-api-key` + `anthropic-version` |
| Gemini native | `x-goog-api-key` or `?key=` |

With the Anthropic SDK, change only the base URL (the SDK appends `/v1/messages` itself):

```python
import anthropic

client = anthropic.Anthropic(api_key="sk-lazu-...", base_url="https://api.lazu.ai")
message = client.messages.create(
    model="claude-opus-5-5",
    max_tokens=512,
    messages=[{"role": "user", "content": "Hello"}],
)
print(message.content[0].text)
```

Gemini's native API goes through `generateContent`:

```bash
curl https://api.lazu.ai/v1beta/models/gemini-3.8-flash:generateContent \
  -H "x-goog-api-key: $LAZU_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"contents": [{"parts": [{"text": "Hello"}]}]}'
```

## 5. What these four models cost today

The cards below read the live catalog, so they change when prices do:

Model: `gpt-5.6-sol` — Stable $5.00 / $30.00 · Discount $0.90 / $5.40 — https://lazu.ai/models/openai/gpt-5.6-sol

Model: `claude-opus-5-5` — Stable $4.00 / $20.00 · Discount $1.72 / $8.60 — https://lazu.ai/models/anthropic/claude-opus-5-5

Model: `gemini-3.8-flash` — Stable $0.75 / $3.75 · Discount $0.1875 / $0.9375 — https://lazu.ai/models/google/gemini-3.8-flash

Model: `deepseek-v4-pro` — Stable $0.825 / $2.48 — https://lazu.ai/models/deepseek/deepseek-v4-pro

## 6. Every call reconciles

Every response carries an `X-Request-Id` header. Find the request by that ID under Usage in the console to see input and output tokens, the lane it took and what it cost, or call `GET /api/usage/requests/{request_id}` with the same key to check it from code.

## Next

- [Quickstart](https://lazu.ai/docs/quickstart): the full setup
- [Pricing and lanes](https://lazu.ai/docs/models/pricing): how the stable and discount lanes differ
- [Model catalog](/models): every available model and its price