Rent a GPU

gpt-oss 20B hardware requirements

20.9B parameters (3.6B active, mixture of experts). Needs about 13.5 GB at Q4_K_M with 8K context.

Parameters21B3.6B active (MoE)
Q4_K_M · 8K13.5 GBthe usual download
FP16 · 8K42.6 GBfull precision
Runs well on33/51GPUs and Macs we track

Memory needed at each quantization

Weights, KV cache and runtime overhead, at 8K and 32K tokens of context.

QuantQuality8K context32K context
FP16Full quality42.6 GB43.2 GB
Q8_0Near-lossless23 GB23.6 GB
Q6_KExcellent17.9 GB18.5 GB
Q5_K_MVery good15.7 GB16.3 GB
Q4_K_MGood, most popular13.5 GB14.1 GB
Q3_K_MNoticeably degraded11 GB11.6 GB
Q2_KHeavily degraded8.6 GB9.2 GB

Run it with Ollama

Downloads Q4_K_M by default.

ollama run gpt-oss:20b

Model card

openai/gpt-oss-20b on Hugging Face: licence, prompt format and GGUF builds.

Run it in the cloud

Run gpt-oss 20B on a rented GPU today.

It needs about 13.5 GB at Q4_K_M: a rented RTX 3090 / RTX 4090 runs it for about $0.18 to $0.89 an hour.

RunPodRent a GPU on RunPodVast.aiRent a GPU on Vast.ai

Every GPU and Mac checked

Sorted by memory. Bars show how much of each card gpt-oss 20B fills at the best quant that fits.

33 Runs well

NameMemory usedBest quantSpeed

12 Fits with heavy quantization

NameMemory usedBest quantSpeed

6 Too big

NameMemory usedBest quantSpeed

Other gpt-oss models