Blog · 2026-07-08

Free GPT-OSS API: OpenAI's Open Model at No Cost

Looking for a free GPT-OSS API? GPT-OSS is OpenAI's open-weight model family, and you can run it right now through the DreamPrompting gateway with no billing and no credit card. All you need is a free API key, which takes seconds to create: sign in with Google on the account page and copy it. This post explains what GPT-OSS actually is, why it matters, and how to send your first request in about two minutes using either curl or the standard Python OpenAI SDK.

What is GPT-OSS?

GPT-OSS is OpenAI's release of open-weight language models, published so that anyone can download, host, and run them. The family includes larger and smaller checkpoints such as gpt-oss-120b and gpt-oss-20b. Being open-weight means the model parameters are freely available, so providers can serve them at low or zero cost, and you are not locked to a single vendor.

One important point of honesty: GPT-OSS is not the same thing as the hosted GPT-4 or ChatGPT products. Those remain closed, paid services. GPT-OSS is a separate open-weight release. It is strong at general chat, reasoning, coding help, and summarization, but you should not expect it to be identical to the flagship paid models. For a huge range of everyday tasks it is more than good enough, and the fact that it is free changes the economics entirely.

Why a free GPT-OSS API is interesting

Because the weights are open, fast inference providers like Groq and Cerebras serve GPT-OSS at no cost to developers. DreamPrompting puts an OpenAI-compatible layer in front of those providers and adds automatic failover, so if one backend is busy your request is routed to another. You get a single stable endpoint, and you do not have to manage keys for several services.

Base URL and endpoints

The gateway base URL is https://dreamprompting.com/api/v1. You send chat requests to POST /api/v1/chat/completions and you list available models with GET /api/v1/models. Full details live in the API docs, and you can browse the current lineup on the models page.

Find the live gpt-oss model id

Provider availability shifts over time, so the exact model id for GPT-OSS can change. The safest approach is to list the models and look for anything containing gpt-oss. A typical pinned id looks like groq/openai/gpt-oss-120b, but always confirm against the live list first.

curl https://dreamprompting.com/api/v1/models \
  -H "Authorization: Bearer YOUR_API_KEY"

Scan the returned ids for a gpt-oss entry and copy the exact string. If you would rather not pin anything, you can pass "auto" and let the gateway choose a healthy free model for you.

Call the free GPT-OSS API with curl

This request uses "auto" so it keeps working even as providers change. Swap in a pinned gpt-oss id once you have confirmed it from the models list.

curl https://dreamprompting.com/api/v1/chat/completions \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "auto",
    "messages": [
      {"role": "user", "content": "Explain what open-weight models are in one sentence."}
    ]
  }'

Call it with the Python OpenAI SDK

Point the standard OpenAI SDK at the gateway base URL. The code below first lists the models to discover the live gpt-oss id, then sends a chat request. If no gpt-oss id is found, it falls back to "auto".

from openai import OpenAI

client = OpenAI(
    base_url="https://dreamprompting.com/api/v1",
    api_key="YOUR_API_KEY",
)

# Find the live gpt-oss model id
model_id = "auto"
for m in client.models.list().data:
    if "gpt-oss" in m.id:
        model_id = m.id
        break

resp = client.chat.completions.create(
    model=model_id,
    messages=[
        {"role": "user", "content": "Give me three ideas for a weekend project."}
    ],
)
print(resp.choices[0].message.content)

To stream tokens as they arrive, add stream=True to the create call and iterate over the response. Because the interface matches OpenAI, most existing code works with only the base URL and key changed. Prefer a chat window instead? Try the models directly in the chat playground.

FAQ

Is this the same as GPT-4? No. GPT-OSS is OpenAI's open-weight model family, not the hosted GPT-4 or ChatGPT products. It is a separate, freely available release that providers can run at low cost. It is capable for everyday tasks, but it is not the paid flagship model.

Is it really free? Yes. The free GPT-OSS API through this gateway costs nothing and never asks for a card. You do need a key, and it is free: sign in with Google at /account, copy it, and pass it as Bearer YOUR_API_KEY. The current limits are 100 requests per minute and a rolling 24 hour account quota of 5,000 requests and 500,000 tokens, which suits testing and small apps.

How do I find the exact model id? Call GET /api/v1/models and look for an id containing gpt-oss, for example groq/openai/gpt-oss-120b. Availability shifts, so check the live list rather than hardcoding an id that may retire.

Start building

A free GPT-OSS API removes the friction of billing and quota negotiations so you can prototype in minutes. Browse the current lineup on the models page, then read the API docs for streaming, parameters, and failover details. Copy the snippets above and send your first request today.

Related