Qwen2.5 14B Instruct on the NVIDIA RTX 3090

Best measured: 56.2 tokens/s generation and 1,950 tokens/s prompt processing (Q4_K_M, llamafile).

For comparison, our size-scaling estimate for this combination is 73.0 tok/s.

Every run

Quant	Backend	Generate tok/s	Prompt tok/s	Test	Build / by / notes	Source
Q4_K_M	llamafile	56.2	1,950	LocalScore avg	LocalScore rank 16/418	LocalScore (localscore.ai)

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 3090 · Qwen2.5 14B Instruct on other hardware