changelog•4 min read

zorveus/gpt-oss-120b is Live on Zorveus

Zorveus Engineering•15 September 2026

zorveus/gpt-oss-120b is Live on Zorveus

OpenAI's open-weight reasoning model zorveus/gpt-oss-120b is live across all Zorveus Edge gateways, official SDKs, and connected application endpoints.

The model uses a sparse Mixture-of-Experts architecture with 117 billion total parameters and activates 5.1 billion parameters per token. It provides a 131,072-token context window with support for configurable reasoning effort, chain-of-thought visibility, and native structured outputs. Zorveus hosts zorveus/gpt-oss-120b with MXFP4 precision to provide high token throughput and low time-to-first-token latency.

# Specifications & architecture

Metric	Specification
Model ID	`zorveus/gpt-oss-120b`
Context Window	131,072 tokens (131K context)
Maximum Output	131,072 tokens
Architecture	Sparse Mixture-of-Experts (117B total, 5.1B active per token)
Precision	MXFP4 (4.25 bits per parameter)
License	Apache 2.0
Modalities	Text, Code, Structured JSON
Reasoning Controls	Configurable reasoning effort (`low`, `medium`, `high`)
Billing	Automatic deduction from Organization Wallet or Product-User grants

# Configuring reasoning effort

zorveus/gpt-oss-120b lets you adjust reasoning intensity directly in request parameters. You can tune the model for low-latency completions or set it to deliberate across multi-step logic paths.

The model accepts three effort levels:

low: Emits fewer reasoning tokens for rapid responses and simple extraction tasks.
medium: Standard reasoning depth for general coding and analysis tasks.
high: Deep chain-of-thought deliberation for multi-step math, architectural audits, and complex planning.

To set the reasoning level, pass reasoning_effort in your request body or client options.

# Using zorveus/gpt-oss-120b in Zorveus

You can start making calls to zorveus/gpt-oss-120b using the official Zorveus SDKs, standard OpenAI client libraries, or direct HTTP requests. Select your integration method from the tabs below:

Method:

import { Zorveus } from "@zorveus/sdk";

const client = new Zorveus({
  apiKey: process.env.ZORVEUS_API_KEY,
});

const response = await client.chat.completions.create({
  model: "zorveus/gpt-oss-120b",
  messages: [
    {
      role: "user",
      content: "Explain distributed consensus in two concise paragraphs.",
    },
  ],
  reasoning_effort: "medium",
  max_tokens: 2048,
});

console.log(response.choices[0].message.content);

import os
from zorveus import Zorveus

client = Zorveus(api_key=os.environ["ZORVEUS_API_KEY"])

response = client.chat.completions.create(
    model="zorveus/gpt-oss-120b",
    messages=[
        {
            "role": "user",
            "content": "Explain distributed consensus in two concise paragraphs.",
        }
    ],
    reasoning_effort="medium",
    max_tokens=2048,
)

print(response.choices[0].message.content)

curl https://api.zorveus.com/v1/chat/completions \
  -H "Authorization: Bearer $ZORVEUS_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "zorveus/gpt-oss-120b",
    "messages": [
      {
        "role": "user",
        "content": "Explain distributed consensus in two concise paragraphs."
      }
    ],
    "reasoning_effort": "medium",
    "max_tokens": 2048
  }'

import os
from openai import OpenAI

client = OpenAI(
    base_url="https://api.zorveus.com/v1",
    api_key=os.environ["ZORVEUS_API_KEY"],
)

response = client.chat.completions.create(
    model="zorveus/gpt-oss-120b",
    messages=[
        {"role": "user", "content": "Explain distributed consensus in two concise paragraphs."}
    ],
    reasoning_effort="medium",
    max_tokens=2048,
)

print(response.choices[0].message.content)

import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://api.zorveus.com/v1",
  apiKey: process.env.ZORVEUS_API_KEY,
});

const response = await client.chat.completions.create({
  model: "zorveus/gpt-oss-120b",
  messages: [
    { role: "user", content: "Explain distributed consensus in two concise paragraphs." },
  ],
  reasoning_effort: "medium",
  max_tokens: 2048,
});

console.log(response.choices[0].message.content);

# Real-time streaming

For conversational applications, terminal agents, and low time-to-first-token loops, zorveus/gpt-oss-120b streams tokens over Server-Sent Events (SSE):

Method:

import { Zorveus } from "@zorveus/sdk";

const client = new Zorveus({
  apiKey: process.env.ZORVEUS_API_KEY,
});

const stream = await client.chat.completions.create({
  model: "zorveus/gpt-oss-120b",
  messages: [
    {
      role: "user",
      content: "Write a high-concurrency event bus in Go.",
    },
  ],
  stream: true,
  reasoning_effort: "low",
});

for await (const chunk of stream) {
  process.stdout.write(chunk.choices[0]?.delta?.content || "");
}

import os
from zorveus import Zorveus

client = Zorveus(api_key=os.environ["ZORVEUS_API_KEY"])

stream = client.chat.completions.create(
    model="zorveus/gpt-oss-120b",
    messages=[
        {"role": "user", "content": "Write a high-concurrency event bus in Go."}
    ],
    stream=True,
    reasoning_effort="low",
)

for chunk in stream:
    print(chunk.choices[0].delta.content or "", end="", flush=True)

curl -N https://api.zorveus.com/v1/chat/completions \
  -H "Authorization: Bearer $ZORVEUS_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "zorveus/gpt-oss-120b",
    "messages": [
      {
        "role": "user",
        "content": "Write a high-concurrency event bus in Go."
      }
    ],
    "stream": true,
    "reasoning_effort": "low"
  }'

import os
from openai import OpenAI

client = OpenAI(
    base_url="https://api.zorveus.com/v1",
    api_key=os.environ["ZORVEUS_API_KEY"],
)

stream = client.chat.completions.create(
    model="zorveus/gpt-oss-120b",
    messages=[
        {"role": "user", "content": "Write a high-concurrency event bus in Go."}
    ],
    stream=True,
    reasoning_effort="low",
)

for chunk in stream:
    print(chunk.choices[0].delta.content or "", end="", flush=True)

import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://api.zorveus.com/v1",
  apiKey: process.env.ZORVEUS_API_KEY,
});

const stream = await client.chat.completions.create({
  model: "zorveus/gpt-oss-120b",
  messages: [
    { role: "user", content: "Write a high-concurrency event bus in Go." },
  ],
  stream: true,
  reasoning_effort: "low",
});

for await (const chunk of stream) {
  process.stdout.write(chunk.choices[0]?.delta?.content || "");
}

# Workspace controls & routing

When deploying zorveus/gpt-oss-120b into production applications, Zorveus gives you full operational control over usage and billing:

Per-user attribution: Pass a user parameter with every request to track usage and costs by client ID, team, or internal feature.
Hard spend caps: Configure period limits on your workspace API keys so usage cuts off before unexpected overages occur.
Bring-Your-Own-Key (BYOK): Route traffic through your own provider endpoints, with automatic failover to the Zorveus gateway when upstream rate limits hit.

zorveus/gpt-oss-120b is available immediately across all Zorveus accounts.

Discuss Online

Previouszorveus/qwen3.8-27b is Live on Zorveus NextJev is live on Zorveus