gpt-oss 120B › NVIDIA RTX PRO 6000 Blackwell Max-Q

gpt-oss 120B on the NVIDIA RTX PRO 6000 Blackwell Max-Q

Best measured: 184 tokens/s generation and 5,449 tokens/s prompt processing (MXFP4, CUDA, FA on).

For comparison, our size-scaling estimate for this combination is 41.1 tok/s.

Every run

Quant	Backend	Generate tok/s	Prompt tok/s	Test	Build / by / notes	Source
MXFP4	CUDA FA on	171	4,494	tg128	ggerganov llama-bench model column: gpt-oss 120B MXFP4 MoE	llama.cpp discussion #15396
MXFP4	CUDA FA on	197	6,404	tg128	SoftwareRenderer llama-bench model column: gpt-oss 120B MXFP4 MoE	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 PRO 6000 Blackwell Max-Q · gpt-oss 120B on other hardware