# gpt-oss-120b

## Best for

Low-cost reasoning, extraction, and tool use over long text

## When not to use

Image, audio, or video. Strict JSON schema output is not enabled yet. Capacity is a single GPU: requests above it get a 429 with Retry-After.

## Overview

gpt-oss-120b is OpenAI's open-weight mixture-of-experts model, with 117B parameters and 5.1B active per token, released under Apache 2.0. It reasons with adjustable effort (low, medium, or high), calls tools, and has a 131,072-token context window. We serve the original MXFP4 weights, with no further quantization, from a dedicated GPU in the US.

| | |
| --- | --- |
| API identifier | `openai/gpt-oss-120b` |
| Developer | OpenAI |
| Context window | 131,072 tokens |
| Supports | text, reasoning, tool-use |
| Status | Serving |

## Rates

USD per 1M tokens, on-demand tier.

| Input | Cached input | Output |
| --- | --- | --- |
| $0.10 | $0.10 | $0.48 |

## Call it

Put `openai/gpt-oss-120b` in the `model` field. The same request works against the Anthropic Messages API — see [API Compatibility](https://inference.boundless.network/docs/api-compatibility).

```bash
curl https://api.inference.boundless.network/v1/chat/completions \
  -H "Authorization: Bearer $BOUNDLESS_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "openai/gpt-oss-120b",
    "messages": [{
      "content": "Say hello",
      "role": "user"
    }]
  }'
```

Every model we serve is listed at [Models](https://inference.boundless.network/models); the rate card in full is on [Pricing](https://inference.boundless.network/docs/pricing).