On our own GPUs
gpt-oss-120b API
OpenAI’s own open-weight reasoning model, under Apache 2.0.
Context
128K
Max output
32K
Serving precision
BF16 / MXFP4
Licence
Apache-2.0
gpt-oss-120b is OpenAI’s open-weight reasoning model, released under Apache 2.0 with a 128K context window. It uses a mixture-of-experts architecture with a small active parameter count, which is why it fits on a single high-end accelerator and why its unit cost lands so low. For bulk, stable, inexpensive inference it remains one of the most pragmatic choices in the open-weight camp.
Open weights, deployed in facilities we build out and lease directly, operated by us. We resell nobody else’s account quota.
Specifications
- Model ID
- gpt-oss-120b
- Published by
- OpenAI
- How it is served
- On our own GPUs
- Licence
- Apache-2.0
- Total parameters
- 120B MoE
- Active parameters
- 5.1B
- Context
- 128K
- Max output
- 32K
- Serving precision
- BF16 / MXFP4
- Input modalities
- Text
- Released
- 2025-08-05
- Price
- See pricing
What it is good at
- A reasoning model that stays cheap
- It emits chain-of-thought yet keeps unit cost low thanks to a small active parameter count. For work that needs some reasoning depth at high throughput, the economics are hard to beat.
- Apache 2.0, no strings
- No restrictions on commercial use, modification or redistribution, and no user-count clause. We can run it legally in our own facilities, and you can stand up the same weights yourself as a control.
- A mature ecosystem
- Released early and widely adopted, so quantisation recipes, serving frameworks and fine-tuning toolchains are all well trodden. Fewest surprises if you want a private deployment or a side-by-side comparison.
Where it fits
- High-volume classification, extraction and rewriting
- Pipelines that need some reasoning depth on a tight budget
- Local or edge deployments as a control or fallback
- A baseline for evaluating open-model capability
Three steps
Change one line of base_url, put the model ID below in the `model` field, and leave the rest of your code alone.
model: "gpt-oss-120b"
curl https://api.nexaix.net/v1/chat/completions \
-H "Authorization: Bearer $NEXAIX_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "gpt-oss-120b",
"stream": true,
"messages": [
{ "role": "user", "content": "Summarise this contract." }
]
}'import os
from openai import OpenAI
client = OpenAI(
# the only new line
base_url="https://api.nexaix.net/v1",
api_key=os.environ["NEXAIX_API_KEY"],
)
# everything below is your existing code
stream = client.chat.completions.create(
model="gpt-oss-120b",
stream=True,
messages=[
{"role": "user", "content": "Summarise this contract."}
],
)
for chunk in stream:
print(chunk.choices[0].delta.content or "", end="")import OpenAI from "openai";
const client = new OpenAI({
baseURL: "https://api.nexaix.net/v1", // the only new line
apiKey: process.env.NEXAIX_API_KEY,
});
// everything below is your existing code
const stream = await client.chat.completions.create({
model: "gpt-oss-120b",
stream: true,
messages: [
{ role: "user", content: "Summarise this contract." }
],
});
for await (const chunk of stream) {
process.stdout.write(chunk.choices[0]?.delta?.content ?? "");
}About this model
Is gpt-oss the same thing as GPT-5?
No. gpt-oss is a separate open-weight series whose weights can be downloaded and deployed freely; the GPT-5 family is proprietary and reachable only through an official commercial channel. Different capability tiers and different purposes.
Is a 128K window enough?
For the vast majority of production work, yes. When you need more, GLM-5.2, DeepSeek V4 and Kimi K3 are all on the fleet with 1M-token windows.
What precision do you serve?
The precision we currently serve is published on this page. gpt-oss also ships an MXFP4 quantised build; we will not silently swap to a lower-precision variant — that is exactly the "silent downgrade" we refuse to do.
Also worth a look
- On our own GPUsKimi K32.8T mixture-of-experts, a 1M-token window, and vision built into the base.kimi-k3
- On our own GPUsGLM-5.2750B sparse MoE, a 1M window, MIT licensed — tuned hard for software engineering.glm-5.2
- On our own GPUsDeepSeek V4 Pro1.6 trillion parameters, thinking and non-thinking modes, MIT licensed.deepseek-v4-pro
Run your own eval on it
Take the trial credit, run the evals you already have against this model, and then decide whether it gets your production traffic.
