Qwen3.8 Flash Next › NVIDIA RTX PRO 6000 Blackwell
Qwen3.8 Flash Next on the NVIDIA RTX PRO 6000 Blackwell
Qwen3.8 Flash Next runs at about 103 tokens/s on the NVIDIA RTX PRO 6000 Blackwell (measured, IQ3_S, CUDA, single-stream llama-server decode).
Best measured: 103 tokens/s generation (IQ3_S, CUDA). The source gives no prompt-processing figure.
Every run
| Quant | Backend | Generate tok/s | Prompt tok/s | Test | Build / by / notes | Source |
|---|---|---|---|---|---|---|
| IQ3_S | CUDA | 103 | – | decode, 4K prompt, 256 tokens | 3cf03257 Every layer on the card, 300 W power cap, median of three, greedy, 4,038-token prompt; no draft head (MTP off). | strata2signal: Flash-Next on three cards |
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. "decode, llama-server" = single-stream generation speed measured against a running llama-server (no speculative decoding).
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 · Qwen3.8 Flash Next on other hardware
Embed this result
HTML
<a href="https://rigbench.pages.dev/model/qwen3-8-flash-next/on/nvidia-rtx-pro-6000-blackwell/"><img src="https://rigbench.pages.dev/badge/qwen3-8-flash-next--nvidia-rtx-pro-6000-blackwell.svg" alt="Qwen3.8 Flash Next · NVIDIA RTX PRO 6000 Blackwell: 103 tok/s (IQ3_S, measured)"></a>
Markdown (GitHub README, forums)
[](https://rigbench.pages.dev/model/qwen3-8-flash-next/on/nvidia-rtx-pro-6000-blackwell/)