Llama 2 7B on the NVIDIA RTX 4070

Best measured: 92.3 tokens/s generation and 3,179 tokens/s prompt processing (Q4_0, Vulkan, FA off).

Every run

Quant	Backend	Generate tok/s	Prompt tok/s	Test	Build / by / notes	Source
Q4_0	Vulkan FA off	92.3	3,179	tg128	9a48399	llama.cpp Vulkan scoreboard (#10879)
Q4_0	Vulkan FA on	91.5	4,294	tg128	9a48399 coopmat2	llama.cpp Vulkan scoreboard (#10879)

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 4070 · Llama 2 7B on other hardware