Can I Run GPT-OSS on 48 GB system RAM?

作者： Jakub Rusinowski · 最后更新： 2026年8月15日

Yes — comfortably

Yes, comfortably — GPT-OSS 20B at Q8_0 needs about 23.4 GB of the 38.4 GB usable on 48 GB system RAM, leaving ~15 GB spare and running at ~16.1 tok/s (estimated), with room for about 131,072 tokens of context.

Confidence: high · Recommended quantization: Q8_0 · Estimated speed: ~16.1 tok/s

Buy This HardwareAMD Radeon RX 7900 XTX 24GB — 24 GB VRAM · 355 W board power Deploy in the Cloud NowRTX 4090 on RunPod — from $0.34/hr · rate checked 2026-07

or compare on Vast.ai from $0.35/hr (typical low · varies)

As an Amazon Associate we earn from qualifying purchases. Cloud GPU links are referral links — we may earn a commission at no extra cost to you.

48 GB system RAM — what it gives a model

Usable memory for models	38.4 GB
Memory bandwidth	90 GB/s

GPT-OSS on 48 GB system RAM: memory by quantization

Quant	Memory needed	Fits 38.4 GB?	Max context	Est. speed	Download
F16	43 GB	✗ No	—	—	41.8 GB
Q8_0	23.4 GB	✓ Yes	128K	~16.1 tok/s	22.2 GB
Q6_K	18.3 GB	✓ Yes	128K	~20.3 tok/s	17.1 GB
Q5_K_M	16 GB	✓ Yes	128K	~23.1 tok/s	14.8 GB
Q4_K_M	13.8 GB	✓ Yes	128K	~26.6 tok/s	12.6 GB
Q3_K_M	10.1 GB	✓ Yes	128K	~35.6 tok/s	8.9 GB
Q2_K	8.1 GB	✓ Yes	128K	~43.8 tok/s	6.9 GB

Which GPT-OSS sizes fit

Variant	Needs at Q4_K_M	Fits?	Est. speed
GPT-oss 120B	71.9 GB	✗ Too large	—
GPT-OSS 20B	13.8 GB	✓ Fits	~26.6 tok/s

What to watch out for

1 larger variant of GPT-OSS does not fit and would need CPU offload or different hardware.
These figures assume CPU-only inference. Any discrete GPU, even an 8 GB one, will be several times faster for models that fit in its VRAM.

Recommended setup

llama.cpp (CPU build) or Ollama — both run without a GPU

How these numbers are calculated

Memory need = quantized weights + KV cache at 8,192 tokens (f16) + 0.8 GB runtime overhead.
38.4 GB of the 48 GB is treated as usable for model weights (80% — the rest is the OS and running applications).
DDR5-5600 dual channel at 89.6 GB/s peak. CPU decode is assumed to sustain 35% of that peak, because CPU inference is not purely bandwidth-bound — it also spends real time in compute and thread synchronisation. This figure is an assumption, not a fitted constant: no CPU measurement is in the calibration set.
CPU-only inference: no GPU is assumed. A GPU of any size will beat these figures substantially.
Throughput is a memory-bandwidth roofline estimate, not a measurement. It is labelled "estimated" everywhere it appears.
KV cache is computed from this model's published attention configuration.

FAQ