Quick Start

Deploy your first Qwen3-14B AI Rig and make your first OpenAI-compatible API request in a few minutes.

1. Deploy an AI Rig

Select Qwen/Qwen3-14B-AWQ.
Choose a supported GPU and deploy your Rig.

Tolka provisions the GPU, starts the inference runtime, and loads the model. Cold-start time varies depending on GPU availability and model initialization, but usually takes 3-5 minutes.

Billing Requirements

Make sure your wallet has sufficient credits before deploying.

2. Get an API key

Create a key from the API Keys Dashboard. Keys look like tk-live-... and are shown once — copy it somewhere safe.

Store keys securely

Tolka Edge only stores a SHA-256 hash of your key. If you lose the plaintext, you must roll a new one.

Export your key so the examples below can read it:

export TOLKA_API_KEY="tk-live-xxxxxxxxxxxxxxxx"

3. Set your base URL

Tolka Edge provides an OpenAI-compatible API. Point your client at:

https://api.tolkaedge.com/v1

4. Make your first request

from openai import OpenAI
import os
 
client = OpenAI(
    base_url="https://api.tolkaedge.com/v1",
    api_key=os.environ.get("TOLKA_API_KEY"),
)
 
# Request hits your dedicated Qwen3-14B rig
resp = client.chat.completions.create(
    model="Qwen/Qwen3-14B",
    messages=[{"role": "user", "content": "Explain dedicated GPU inference in one sentence."}],
)
 
print(resp.choices[0].message.content)

import OpenAI from "openai";
 
const client = new OpenAI({
  baseURL: "https://api.tolkaedge.com/v1",
  apiKey: process.env.TOLKA_API_KEY,
});
 
async function main() {
  const resp = await client.chat.completions.create({
    model: "Qwen/Qwen3-14B",
    messages: [{ role: "user", content: "Explain dedicated GPU inference in one sentence." }],
  });
 
  console.log(resp.choices[0].message.content);
}
 
main();

curl https://api.tolkaedge.com/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $TOLKA_API_KEY" \
  -d '{
    "model": "Qwen/Qwen3-14B",
    "messages": [
      {
        "role": "user",
        "content": "Explain dedicated GPU inference in one sentence."
      }
    ]
  }'

5. Streaming (Recommended)

Streaming is recommended for lower perceived latency and incremental responses.

from openai import OpenAI
 
client = OpenAI(
    base_url="https://api.tolkaedge.com/v1",
    api_key="tk-live-xxxxxxxxxxxxxxxx",
)
 
stream = client.chat.completions.create(
    model="Qwen/Qwen3-14B",
    messages=[{"role": "user", "content": "Write a short poem about GPUs."}],
    stream=True,
)
 
for chunk in stream:
    if chunk.choices[0].delta.content:
        print(chunk.choices[0].delta.content, end="", flush=True)

import OpenAI from "openai";
 
const client = new OpenAI({
  baseURL: "https://api.tolkaedge.com/v1",
  apiKey: "tk-live-xxxxxxxxxxxxxxxx",
});
 
async function main() {
  const stream = await client.chat.completions.create({
    model: "Qwen/Qwen3-14B",
    messages: [{ role: "user", content: "Write a short poem about GPUs." }],
    stream: true,
  });
 
  for await (const chunk of stream) {
    process.stdout.write(chunk.choices[0]?.delta?.content || "");
  }
}
 
main();

Next steps

Explore the API Reference.
Learn about the Dedicated GPU lifecycle, shutdown, and billing.