Best GPU for GPT‑OSS 120B (Cloud + Buy Hardware)

Best GPU for GPT‑OSS 120B

Cheapest cloud GPUs and best local hardware to run GPT‑OSS 120B efficiently.

120 GB Recommended VRAM

80 GB Minimum VRAM

120B class Model Size

120GB+ VRAM is ideal for stable inference and larger context.

80GB works for quantized configs with reduced batch size.

Production tier is strongly recommended for long runs.

Best Cloud GPU Offers

Cheapest offers that meet the VRAM requirements for this model.

Provider	GPU	VRAM	$/GPU‑Hr	Availability	Tier	Link
RunPod	MI300X	192.0 GB	$0.50	in_stock	mixed	View
RunPod	H200 NVL	143.0 GB	$0.50	in_stock	mixed	View
Crusoe	A100 80GB	80.0 GB	$1.09	in_stock	production	View
RunPod	A100 40GB	80.0 GB	$1.19	in_stock	mixed	View
Shadeform	A100 80GB	80.0 GB	$1.19	in_stock	mixed	View
Lambda	A100 80GB	80.0 GB	$1.29	request_required	production	View
RunPod	A100 SXM	80.0 GB	$1.39	in_stock	mixed	View
MassedCompute	A100 80GB	80.0 GB	$1.39	in_stock	production	View
FluidStack	A100 80GB	80.0 GB	$1.40	in_stock	production	View
Replicate	A100 80GB	80.0 GB	$1.40	in_stock	unknown	View
Hyperstack	A100 80GB	80.0 GB	$1.49	in_stock	production	View
Hetzner	A100 80GB	80.0 GB	$1.54	request_required	production	View

Best GPUs to Buy

Top local hardware options that meet the minimum VRAM requirement.

Rank	GPU	VRAM	Price	AI Score	$ Value	Buy	Compare
—	NVIDIA H100 PCIe	80 GB	$30,000	100.0	3.33	Amazon eBay

VRAM Hunter