gpt-oss 20B on the NVIDIA RTX 5080

Best measured: 229 tokens/s generation and 10,404 tokens/s prompt processing (MXFP4, CUDA, FA on).

For comparison, our size-scaling estimate for this combination is 177 tok/s.

Every run

Quant	Backend	Generate tok/s	Prompt tok/s	Test	Build / by / notes	Source
MXFP4	CUDA FA on	205	7,477	tg128	ggerganov llama-bench model column: gpt-oss 20B MXFP4 MoE	llama.cpp discussion #15396
MXFP4	CUDA FA on	225	9,558	tg128	PaulKnowler llama-bench model column: gpt-oss 20B MXFP4 MoE	llama.cpp discussion #15396
MXFP4	CUDA FA on	225	9,558	tg128	PaulKnowler llama-bench model column: gpt-oss 20B MXFP4 MoE	llama.cpp discussion #15396
MXFP4	CUDA FA on	225	9,558	tg128	PaulKnowler llama-bench model column: gpt-oss 20B MXFP4 MoE	llama.cpp discussion #15396
F16	CUDA FA on	178	12,995	tg128	PaulKnowler llama-bench model column: gpt-oss 20B F16	llama.cpp discussion #15396

tg128 = generating 128 tokens; pp512 = reading a 512-token prompt (llama-bench). LocalScore averages several prompt and output lengths, so its numbers run lower than tg128.

Check prices

Amazon Newegg

As an Amazon Associate I earn from qualifying purchases.

No card yet?

Try it on a rented GPU (RunPod)

Affiliate link: if you sign up, RigBench may earn a referral credit. It never affects rankings or numbers.

All results for the NVIDIA RTX 5080 · gpt-oss 20B on other hardware