Model APIs

Run open models without managing GPUs.

Use hosted models through an OpenAI-compatible API.Pay for the tokens you use.

Try the playgroundView models & pricing

Models and pricing

Pay for the tokens you use.

Use the Jarvislabs LLM API to access hosted open models. Compare input and output prices, then choose a model to try.

Scroll across the table to compare all token rates.

USD per 1 million tokens. Prices checked September 15, 2026.
ModelInputCached inputOutput
GLM-5.3$0.90$0.15$2.80
GLM-5.3 Flash$0.07$0.014$0.22
DeepSeek V4 Flash$0.07$0.015$0.20
Gemma 4 26B A4B IT$0.05$0.01$0.20
Gemma 4 31B$0.09$0.02$0.30
GPT-OSS 120B$0.04$0.01$0.18
Muse Glimmer 30B$0.20$0.03$0.80
NVIDIA Nemotron 3.5 Lightning 30B A3B BF16$0.06$0.015$0.16
Qwen3.8 27B FP8$0.18$0.04$1.50

Input is the context you send; output is what the model generates. Cached input rates apply to verified cache hits. Sign in for current availability and prices in your billing currency.

Open the model catalog

From prompt to application

Use the OpenAI SDK.

Send Chat Completions requests through an OpenAI-compatible API. Set your Jarvislabs API key, base URL and model ID, then use Python, TypeScript or cURL.

  1. Choose a model

    This example uses GLM-5.3. Browse the catalog for other models.

    Browse models

  2. Create an API key

    Set JARVISLABS_API_KEY in your server environment. Keep the key out of browser code.

    Manage API keys

  3. Make your first request

    Install the SDK, copy the example and run it from your server or terminal.

pip install openai

import os
from openai import OpenAI

client = OpenAI(
    base_url="https://models.jarvislabs.net/v1",
    api_key=os.environ["JARVISLABS_API_KEY"],
)

response = client.chat.completions.create(
    model="glm-5.3",
    messages=[
        {"role": "user", "content": "Say hello in one sentence."},
    ],
)
print(response.choices[0].message.content)

Set JARVISLABS_API_KEY in your environment before running.

The playground

Try a prompt before writing code.

Choose a model, send a prompt and see how it responds. Adjust your request, then export it as code for your application.

Try the playground

Choose your model

Start with a model from the catalog and the prompt you want to test.

Refine your request

Edit your prompt and adjust the available model settings as you try it.

Take it into your app

Export the request as Python, TypeScript or cURL to continue building.

More control

Need your own deployment?

Choose the deployment options that fit your model and serving stack. Managed Endpoints and Serverless both bill for worker runtime.

Managed Endpoints

Deploy a catalog model on a serving stack we maintain. You choose how your deployment scales.

Explore Managed Endpoints

Serverless

Deploy your own model and serving stack with an autoscaling worker pool.

Explore Serverless

Before you start

Common questions.

Practical details for choosing and using this product.

Do I need to rent a GPU first?

No. Model APIs gives you access to hosted models. Choose a model and call it through the API.

Which models can I use?

The catalog on this page includes GLM, DeepSeek, Gemma, GPT-OSS, Muse, Nemotron and Qwen models, with input and output token rates. The dashboard shows current availability and model details.

Can I use the OpenAI SDK?

Yes, for Chat Completions. Set the Jarvislabs base URL, API key and model ID. The quickstart on this page includes Python, TypeScript and cURL examples.

How is API usage charged?

By input and output tokens, at the rates listed for the model you use. The catalog shows these rates separately so you can compare models before making a request.

How is this different from Managed Endpoints and Serverless?

With Model APIs, you call a model that is already hosted. With Managed Endpoints or Serverless, you create your own deployment and pay for worker runtime.

Choose a model and start building.

Try a prompt in the playground, then bring the request into your application.

Try the playground