Radeon PRO W7900
AMD
ExcellentRuns gpt-oss-120b MXFP4 entirely in memory
Model · OpenAI
gpt-oss-120b: VRAM and hardware requirements
At MXFP4 with an 8K context, gpt-oss-120b needs about 71.2 GB: a GPU with 80 GB or more runs it entirely in VRAM.
- Parameters116.8B
- Active5.1B
- Max context128K
- LicenseApache-2.0
Specifications
- Organization
- OpenAI
- Architecture
- gpt_oss
- Released
- August 2025
- Layers
- 36
- Hidden size
- 2880
- Attention heads
- 64
- KV heads
- 8
- Head dimension
- 64
- Vocabulary
- 201,088
- KV cache per 1K tokens (f16)
- 0.04 GB
- Repository
- openai/gpt-oss-120b
Official specData updated 24 days agoSource: Hugging Face config.json + model card
Quantizations and memory
| Quantization | File size | Memory needed | Minimum GPU memory | Quality |
|---|---|---|---|---|
| MXFP4Most used | 59 GB | ~71.2 GB | 80 GB | Reference (unquantized) |
File sizes: not available until imported from the model repository. Memory = weights (estimated from bits per weight) + KV cache + runtime overhead + safety margin. (MXFP4 ≈ 4.9 bpw)
Q4, Q5 or Q8: how to choose a quantization
GPUs that run gpt-oss-120b
| RTX 509032 GB · 64 GB RAM~63.2 GB · 13–24 tok/s estimate | Possible but slow |
|---|---|
| RTX 409024 GB · 64 GB RAM~63.2 GB · 10–20 tok/s estimate | Possible but slow |
| RTX 309024 GB · 64 GB RAM~63.2 GB · 10–20 tok/s estimate | Possible but slow |
| RX 7900 XTX24 GB · 64 GB RAM~63.3 GB · 9.5–19 tok/s estimate | Possible but slow |
| RTX 5060 Ti 16 GB16 GB · 64 GB RAM~63.2 GB · 8.5–17 tok/s estimate | Possible but slow |
| RTX 5070 Ti16 GB · 64 GB RAM~63.2 GB · 8.5–17 tok/s estimate | Possible but slow |
| RTX 508016 GB · 64 GB RAM~63.2 GB · 8.5–17 tok/s estimate | Possible but slow |
| RTX 4060 Ti 16 GB16 GB · 64 GB RAM~63.2 GB · 8.5–16 tok/s estimate | Possible but slow |
| RTX 4070 Ti SUPER16 GB · 64 GB RAM~63.2 GB · 8.5–17 tok/s estimate | Possible but slow |
| RX 9070 XT16 GB · 64 GB RAM~63.3 GB · 8.5–17 tok/s estimate | Possible but slow |
| RTX 3060 12 GB12 GB · 64 GB RAM~63.2 GB · No data | Incompatible |
| RTX 407012 GB · 64 GB RAM~63.2 GB · No data | Incompatible |
| RTX 507012 GB · 64 GB RAM~63.2 GB · No data | Incompatible |
| RTX 40608 GB · 64 GB RAM~63.2 GB · No data | Incompatible |
Macs that run gpt-oss-120b
| M3 Ultra (80-core GPU) · 512 GBMac Studio~63 GB · 70–92 tok/s estimate | Excellent |
|---|---|
| M4 Max (40-core GPU) · 128 GBMacBook Pro 14, MacBook Pro 16~63 GB · 44–60 tok/s estimate | Excellent |
| M3 Max (40-core GPU) · 128 GBMacBook Pro 14, MacBook Pro 16~63 GB · 32–44 tok/s estimate | Excellent |
| M4 Pro (20-core GPU) · 64 GBMac mini, MacBook Pro 14~63 GB · No data | Incompatible |
| M4 Max (32-core GPU) · 36 GBMacBook Pro 14, MacBook Pro 16~63 GB · No data | Incompatible |
| M4 · 32 GBMacBook Air, MacBook Pro 14~63 GB · No data | Incompatible |
| M5 · 32 GBMacBook Pro 14~63 GB · No data | Incompatible |
| M3 · 24 GBMacBook Air, MacBook Pro 14~63 GB · No data | Incompatible |
EstimateData updated 24 days agoSource: Run LLM Locally engine
Limits in plain words
gpt-oss-120b is a mixture-of-experts model: all 116.8B parameters must be in memory, but only 5.1B are read per token, so it generates much faster than a dense model of the same size.
The KV cache grows with context: at 32K it adds about 1.1 GB on top of the weights.
With under 8 GB of VRAM, use a smaller quantization or a smaller model of the same family.
Where to buy
Prices and links follow the market you choose.
NVIDIA DGX Spark (GB10)
Unified memory PC
ExcellentRuns gpt-oss-120b MXFP4 entirely in memory
No current offer imported for this market: the link opens a marketplace search.
Frequently asked questions
How much VRAM does gpt-oss-120b need?
About 71.2 GB at MXFP4 with an 8K context, including the KV cache and runtime overhead. At MXFP4 it needs about 71.2 GB. A longer context adds more: about 1.1 GB of KV cache at 32K.
Can gpt-oss-120b run on a Mac?
Yes on a Mac whose GPU can use at least 63 GB, e.g. M4 Max (40-core GPU) 128 GB. macOS lets the GPU use about two thirds to three quarters of unified memory by default.
How fast is gpt-oss-120b?
On a RTX 4090, we estimate 10–20 tok/s at MXFP4. This is an estimate from memory bandwidth (confidence low) until benchmarks are imported.