OpenAI: gpt-oss-120b

GPT-OSS 120B is OpenAI's larger open-weight reasoning model, using a mixture-of-experts design. It is suited to problem solving and tool-driven workflows where an open model's inspectability and customization potential are valuable.

Compare models

gpt-oss-120b

AvailableOpenAIChat

Input / Output: $0.15 / $0.60
Context: 128K
Max output: 64K
Modalities: Chat
Capabilities: Tool usePrompt CacheReasoning

About gpt-oss-120b

gpt-oss-120b is OpenAI's larger open-weight reasoning model, released under the Apache 2.0 license. It is a mixture-of-experts model with about 117 billion parameters, of which roughly 5.1 billion are active per token, sized to run on a single H100 GPU. Compared with gpt-oss-20b it trades a heavier footprint for stronger reasoning, and it takes text in and gives text out.

Where it works well

Open weights under Apache 2.0, so you can inspect, fine-tune, and self-host the same model you call through an API.
Reasoning effort is adjustable, letting you spend more thinking on hard problems and less on quick lookups.
The full chain of thought is visible, which helps when you need to debug why the model reached an answer.
Native function calling, web search, code execution, and structured outputs suit agent-style workflows.

When to choose another model

It reads and writes text only, so documents that depend on charts or scans need a different model.
Its training data ends in mid-2024, so recent events need a search tool or supplied context.
For the hardest closed-model reasoning tasks, frontier proprietary models usually remain the stronger pick.

Getting started

Create API key
Create a key in Console, then use it with every model on the platform.
02
Send your first request
Copy the example for your language and run it against the endpoint.
Text to textResponses API
POST/v1/responses
API format:
```
curl https://api.tokenlab.sh/v1/responses \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer sk-xxx" \
  -d '{
    "model": "gpt-oss-120b",
    "input": "Hello!"
  }'
```

Pricing

The Verified price applies to Verified, which costs less on most models. The Official price is the model maker's published price and applies to the more reliable Official route. Auto bills the route that completes the request.

Default spec

per 1M tokens

Official price: Input $0.15 / Output $0.60 / Cache read $0.075
Verified price: —
Discount: —

Prompt cache pricing

Cache read

Official price: $0.075
Verified price: —
Discount: —

	Official priceper 1M tokens	Verified priceper 1M tokens	Discount
Default spec	Input $0.15 / Output $0.60 / Cache read $0.075	—	—
Prompt cache pricing
Cache read	$0.075	—	—

Usage & activity

Success rate is the share of requests that completed. Latency is how long a full response takes; P95 means 95% of requests finished within that time.

Usage & availability

Last 24 hours

Requests
Success rate
P95 latency
Total tokens

Model performance

Last updated Oct 12, 2026, 09:27 PM UTC

Metrics appear once privacy and data volume thresholds are met.

Data is based on aggregate user requests, excluding status checks.

Open in Console

Open gpt-oss-120b in Console with a prompt ready to edit or send.

Help me try gpt-oss-120b with a short message at /v1/responses. Show the reply, latency, and cost.

Use cases

Best for

Reasoning

Self-hostable reasoning backend
Prototype against the hosted model, then move the same weights onto your own H100-class hardware when data residency or cost control requires it.
Tool-calling agents
Drive multi-step workflows that call functions, search the web, and run code, with adjustable effort per step.
Domain fine-tuning
Start from open weights to adapt a reasoning model to a specialist vocabulary such as legal, medical, or internal engineering text.
Auditable reasoning
Keep the visible reasoning trace alongside answers for review in regulated or high-stakes analysis.

Prompt examples

A function returns wrong totals when the input list is empty. Reason through the cause, then give a corrected version with two test cases.

Plan the steps to reconcile two CSV exports of customer invoices, and say which tool calls you need at each step.

Solve this scheduling puzzle with six tasks and three dependencies, and show the reasoning before the final order.

This model has conditional pricing. Monthly token totals alone cannot produce a reliable estimate; use the detailed pricing for the request specification, cache, and applicable time window.

FAQ

What kind of model is gpt-oss-120b?

It is OpenAI's larger open-weight language model for reasoning and tool use. It uses a mixture-of-experts design with about 117 billion total and 5.1 billion active parameters, ships under Apache 2.0, and accepts and produces text.

How is gpt-oss-120b different from gpt-oss-20b?

The 120b model is the larger, more capable sibling meant for single-H100 deployments, while gpt-oss-20b has about 21 billion parameters with 3.6 billion active and targets low-latency, local, or specialized use. Pick 120b when answer quality on reasoning matters more than footprint.

Can I fine-tune or self-host gpt-oss-120b?

Both. The weights are published on Hugging Face under Apache 2.0, which has no copyleft restriction, so you can fine-tune them and run them on your own infrastructure. OpenAI sizes the model for a single H100 GPU.

Does gpt-oss-120b support function calling and structured output?

Yes. OpenAI documents function calling, structured outputs, streaming, and tool use such as web search, code execution, and MCP. Reasoning effort can be adjusted to match your latency target.

Which inputs does gpt-oss-120b accept?

No images or audio: both input and output are text. Keep gpt-oss-120b for text reasoning and tool-driven steps, and pick a multimodal model for visual material.

How much does gpt-oss-120b cost?

On TokenLab, gpt-oss-120b costs Input $0.15 / Output $0.60 per 1M tokens. The pricing table above shows the full breakdown. Rates depend on the billing unit, specification, and usage. Compare matching conditions in the model's detailed pricing; a single rate does not determine the total cost.

What are the context window and output limit of gpt-oss-120b?

gpt-oss-120b accepts up to 131,072 tokens of context and returns up to 65,536 tokens in one response.

Which endpoint should gpt-oss-120b use?

Use https://api.tokenlab.sh/v1/responses for gpt-oss-120b. The request example below shows the matching code shape.

Which operations does gpt-oss-120b support?

gpt-oss-120b supports Text to text. Select an operation above to see its endpoint and request example.