Arrakis / Models / gpt-oss-20b
gpt-oss-20b GPU requirements
gpt-oss-20b has 21B parameters. To serve it, plan for about 50 GB of GPU memory in 16-bit, 25 GB in 8-bit or 13 GB in 4-bit, including about 20% for the KV cache and activations.
It is a mixture of experts: 32 experts, 4 of them active for each token. All of them stay in GPU memory, so plan memory for all 21B parameters.
What it takes to serve it
GPU memory to run it for inference, with the smallest GPU setup that holds it and the lowest price on Arrakis right now.
| Precision | Memory and the smallest setup |
|---|---|
| 16-bit | 50 GB: one H100 or A100, H100 from $1.65 per GPU-hour |
| 8-bit | 25 GB: one RTX 5090, from $0.58 per GPU-hour |
| 4-bit | 13 GB: one RTX 4090 or RTX 3090, RTX 3090 from $0.19 per GPU-hour |
What it takes to fine-tune it
QLoRA trains adapters on a 4-bit frozen model, LoRA on a 16-bit one, and full fine-tuning updates every weight. Figures are before activations, which grow with batch size and sequence length.
| Method | Memory and the smallest setup |
|---|---|
| QLoRA | 15 GB: one RTX 4090 or RTX 3090, RTX 3090 from $0.19 per GPU-hour |
| LoRA | 50 GB: one H100 or A100, H100 from $1.65 per GPU-hour |
| Full fine-tuning | 335 GB: 2 × B200, from $13.58 per hour for all 2 |
About the model
From its card and files on Hugging Face. Memory figures here follow the GPU memory guide.
- Parameters
- 21B
- Experts
- 32
- Experts for each token
- 4
- Context window (tokens)
- 131,072
- Published in
- FP8
- License
- apache-2.0
- Released
- 2025-08-04
- Weights
- openai/gpt-oss-20b
Deploy it on Arrakis
Open Models in the console, choose vLLM, SGLang or llama.cpp, and rent a GPU it fits on. You get an OpenAI-compatible endpoint with its own API key, and root SSH to the machine. Or run it yourself on any machine with enough GPU memory:
vllm serve openai/gpt-oss-20b --tensor-parallel-size 1Questions
Live prices for every GPU are on the GPU prices page.