gpt-oss-20b
The small gpt-oss: o3-mini-class reasoning in a model that fits on a laptop.
About gpt-oss-20b
gpt-oss-20b is the smaller of OpenAI’s two open-weight models. It runs in 16GB of memory — small enough for a high-end laptop — and OpenAI says it performs similarly to o3-mini on common benchmarks, with 71.5% on GPQA Diamond.
It has the same reasoning levels and tool use as the 120b at a fraction of the size. Through an API its appeal is speed and price on simple reasoning steps; for anything hard, the 120b is clearly stronger.
Where it’s strong
Quick reasoning steps inside an agent.
Simple classification and extraction that needs a little thought.
Prototyping on a budget.
Where to be careful
Clearly weaker than gpt-oss-120b on hard reasoning — 71.5% against 80.1% on GPQA Diamond.
Its knowledge stops at June 2024, and it reads text only.
Price
For scale: 1,000 requests of about 3,000 tokens in and 1,000 out cost $0.60 — a long prompt and a page of answer each. Every rate is published and stays the same from one request to the next. See every model.
Specs
gpt-oss-20bCall it
Use the OpenAI SDK you already have — point it at Roar AI and name the model. Read the docs.
from openai import OpenAI
client = OpenAI(base_url="https://api.roar-ai.com/v1", api_key="roar_live_…")
reply = client.chat.completions.create(
model="gpt-oss-20b",
messages=[{"role": "user", "content": "Summarise this contract in plain English."}],
)Compare gpt-oss-20b
Questions
How much does gpt-oss-20b cost?
Input is $0.091 per 1M tokens; output is $0.325 per 1M tokens. For scale, 1,000 requests of about 3,000 tokens in and 1,000 out cost $0.60. The rate is published and does not change from one request to the next.
Can I use gpt-oss-20b alongside other models?
Yes. Every model on Roar AI is on the same API key and the same monthly invoice, so switching from gpt-oss-20b to another model is a change to one word in your code.
Is gpt-oss-20b open source?
Its weights are open under the Apache 2.0 licence, so you could run it yourself. Through Roar AI you call it like any other model, with no servers to manage.