GPT-5.4 nano
OpenAI’s small model for classifying, extracting and ranking at scale.
About GPT-5.4 nano
GPT-5.4 nano is built for the plain, repetitive jobs: sorting text into categories, pulling fields out of documents, ranking results and acting as a small helper inside a bigger agent. It answers without reasoning by default, which keeps it fast, and can be asked to think when a task needs it.
It is not a general assistant. On OSWorld, the computer-use test, it scored 39.0% against 72.1% for its bigger sibling GPT-5.4 mini, and the newer Luna models now cover the same ground with a much longer memory.
Where it’s strong
Classifying and tagging large volumes of text.
Pulling structured fields out of documents.
Ranking and reranking search results.
Simple helper agents.
Where to be careful
Weak at operating software and at multi-step agent work.
Its knowledge stops at August 2025, and its 400K context is well under the newer models’ million.
Price
For scale: 1,000 requests of about 3,000 tokens in and 1,000 out cost $2.13 — 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-5.4-nanoCall 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-5.4-nano",
messages=[{"role": "user", "content": "Summarise this contract in plain English."}],
)Compare GPT-5.4 nano
Questions
How much does GPT-5.4 nano cost?
Input is $0.23 per 1M tokens; output is $1.4375 per 1M tokens. For scale, 1,000 requests of about 3,000 tokens in and 1,000 out cost $2.13. The rate is published and does not change from one request to the next.
Can I use GPT-5.4 nano alongside other models?
Yes. Every model on Roar AI is on the same API key and the same monthly invoice, so switching from GPT-5.4 nano to another model is a change to one word in your code.
Is GPT-5.4 nano open source?
No. OpenAI does not publish the weights, so it is only available through an API like this one.