GPT-OSS — local AI model by OpenAI
Written by Jakub Rusinowski · Last updated
OpenAI's open-weight family released under Apache 2.0 — their first substantial open-source release since GPT-2. Both sizes ship in MXFP4 precision, which is why they fit in far less memory than their parameter counts suggest: the 20B loads in ~14 GB and runs on a 16GB consumer card, while the 120B needs ~65 GB (dual RTX 3090 or a single 80GB A100). Signals OpenAI's strategic shift toward supporting on-premise enterprise deployments.
Variants
The smallest GPT-OSS variant needs about 13 GB of VRAM at Q4_K_M — quantized weights plus framework overhead, before any KV cache.
| Model | VRAM at Q4 | VRAM | Context | Run it |
|---|---|---|---|---|
| GPT-OSS 120B → 120B | ~71.3 GB | 131,072 | ollama run gpt-oss:120b | |
| GPT-OSS 20B → 20B | ~13.4 GB | 128,000 | ollama run gpt-oss:20b |
Memory is quantized weights plus overhead at Q4_K_M, from the same engine as the GPU & VRAM checker.
How to run GPT-OSS locally
Install Ollama, then pull the tag.
ollama run gpt-oss:120bPick a size above for its own VRAM figure, speed estimate and install command.
Licence
Commercial use permitted. No usage restrictions beyond attribution.
Applies to: GPT-OSS 120B, GPT-OSS 20BRecommended GPU
The cheapest catalogued GPU that runs GPT-OSS locally (min 13 GB VRAM) is the AMD Radeon RX 9060 XT 16GB (16 GB).
Can I run GPT-OSS on my GPU?
- GPT-OSS on AMD Radeon RX 7800 XT
- GPT-OSS on AMD Radeon RX 7900 XT
- GPT-OSS on AMD Radeon RX 7900 XTX
- GPT-OSS on AMD Radeon RX 9060 XT 16GB
- GPT-OSS on AMD Radeon RX 9060 XT 8GB
- GPT-OSS on AMD Radeon RX 9070
- GPT-OSS on AMD Radeon RX 9070 XT
- GPT-OSS on AMD Ryzen AI Max+ 395
- GPT-OSS on Apple M1
- GPT-OSS on Apple M1 Ultra
- GPT-OSS on Apple M2
- GPT-OSS on Apple M2 Max
- GPT-OSS on Apple M3
- GPT-OSS on Apple M4 Max
- GPT-OSS on Apple M4 Pro
- GPT-OSS on Apple M5 Max
- GPT-OSS on Intel Arc B570
- GPT-OSS on Intel Arc B580
- GPT-OSS on NVIDIA A100 80GB (PCIe)
- GPT-OSS on NVIDIA DGX Spark
- GPT-OSS on NVIDIA GeForce RTX 3060 (12GB)
- GPT-OSS on NVIDIA GeForce RTX 3070
- GPT-OSS on NVIDIA GeForce RTX 3070 Ti
- GPT-OSS on NVIDIA GeForce RTX 3080 (10GB)
- GPT-OSS on NVIDIA GeForce RTX 3090
- GPT-OSS on NVIDIA GeForce RTX 4060
- GPT-OSS on NVIDIA GeForce RTX 4060 Ti 16GB
- GPT-OSS on NVIDIA GeForce RTX 4070
- GPT-OSS on NVIDIA GeForce RTX 4070 Super
- GPT-OSS on NVIDIA GeForce RTX 4070 Ti
- GPT-OSS on NVIDIA GeForce RTX 4070 Ti Super
- GPT-OSS on NVIDIA GeForce RTX 4080
- GPT-OSS on NVIDIA GeForce RTX 4080 Super
- GPT-OSS on NVIDIA GeForce RTX 4090
- GPT-OSS on NVIDIA GeForce RTX 5060
- GPT-OSS on NVIDIA GeForce RTX 5060 Ti 16GB
- GPT-OSS on NVIDIA GeForce RTX 5060 Ti 8GB
- GPT-OSS on NVIDIA GeForce RTX 5070
- GPT-OSS on NVIDIA GeForce RTX 5070 Ti
- GPT-OSS on NVIDIA GeForce RTX 5080
- GPT-OSS on NVIDIA H100 80GB (PCIe)
- GPT-OSS on NVIDIA RTX PRO 6000 Blackwell
GPT-OSS — frequently asked questions
How much VRAM does GPT-OSS need?
GPT-OSS needs about 13 GB VRAM at Q4_K_M quantization for its smallest variant. Variants: GPT-OSS 120B (71 GB, Q4_K_M); GPT-OSS 20B (13 GB, MXFP4). On Apple Silicon, unified memory counts toward this requirement.
Can I run GPT-OSS on an RTX 4090 (24 GB)?
Yes — GPT-OSS runs on an RTX 4090 (24 GB) and other 24 GB cards such as the RTX 3090. Smaller variants also fit comfortably on 8–16 GB GPUs at Q4_K_M.
What quantization should I use for GPT-OSS?
Q4_K_M is the best balance of quality and VRAM for GPT-OSS in most cases. Choose Q8_0 for near-lossless quality if you have spare VRAM, or smaller quants (Q3/Q2) only when memory is tight.
How do I run GPT-OSS with Ollama?
Install Ollama, then run: ollama run gpt-oss:120b. This downloads GPT-OSS and starts a local, OpenAI-compatible endpoint — no internet connection is needed after the initial download.