## Specs & substance

### Spec sheet

- **Developed by**: OpenAI  
- **Model family**: gpt-oss  
- **Category**: compact  
- **Modality**: Text  
- **Context window**: 128K tokens  
- **Architecture**: MoE  
- **Version**: 20B  
- **License**: Apache 2.0  
- **Pricing**: $0.04 in, $0.20 out · $0.02 cache  
- **Released**: Aug 2025  
- **Endpoint**: parasail-gpt-oss-20b

Compact open-weight model from OpenAI — ultra-cheap, fits on a single GPU.

gpt-oss-20b is part of the gpt-oss family by OpenAI, categorized as a **compact** model. It supports a 128K tokens context window and is built on a MoE architecture.

Key strengths: efficient, open-weight. Parasail serves gpt-oss-20b on a global fleet of current-gen GPUs behind a single OpenAI-compatible endpoint — with per-token pricing, no minimums, and dedicated capacity options when you need guaranteed throughput.

## Integrate

### Drop-in via the OpenAI SDK

Point any OpenAI-compatible chat client at Parasail and change the model name. That's it.

#### Python

```python
from openai import OpenAI

client = OpenAI(
    api_key="YOUR_API_KEY",
    base_url="https://api.parasail.io/v1"
)

response = client.chat.completions.create(
    model="parasail-gpt-oss-20b",
    messages=[
        {"role": "user", "content": "Hello, what can you do?"}
    ],
    stream=True,
    stream_options={"include_usage": True},
    top_p=1,
    max_tokens=1000,
    temperature=1
)

for chunk in response:
    if chunk.choices and chunk.choices[0].delta.content is not None:
        print(chunk.choices[0].delta.content, end="", flush=True)
```

#### Bash

```bash
curl https://api.parasail.io/v1/chat/completions \
  -H "Authorization: Bearer $PARASAIL_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "parasail-gpt-oss-20b",
    "messages": [{"role": "user", "content": "Hello, what can you do?"}],
    "stream": true,
    "max_tokens": 1000
  }'
```

## More models

### Explore the library

- **Model DeepSeek V4 Flash**  
  Ultra-fast, ultra-cheap reasoning model for high-throughput workloads.  
  **$0.14/M in**

- **Model Qwen3.6 35B-A3B**  
  Efficient MoE model with 3B active parameters — great price-to-performance.  
  **$0.15/M in**

- **Model Qwen3.5 35B-A3B**  
  Compact MoE model with strong reasoning at minimal cost per token.  
  **$0.15/M in**

## Run gpt-oss-20b in the platform

Start with free credits — no credit card. Call any model in minutes.