More
Submit build

Builds / RTX Pro 6000 Max-Q (Blackwell, 96 GB)

RTX PRO 6000 Blackwell Max-Q: Qwen3.6 35B-A3B NVFP4 on vLLM at about 261 tok/s single stream, prefill 26.6k tok/s at 8K (URE)

Last updated:

Source: ure.us ↗📏 How measured: reported by the original source; RigBase re-checked the numbers at that source
GPU(s)
RTX Pro 6000 Max-Q (Blackwell, 96 GB)
Amazon · eBay (used) · Newegg
VRAM / unified memory (GB)
96
CPU
not stated
Motherboard
not stated
RAM
not stated
PSU
not stated
Other
Workstation, 300 W cap; vLLM (0.27.1 in the end-to-end run) with the native b12x FP4 MoE kernel, tuned for production
Reported total cost
not stated
Power
300 W card cap (the author notes the card sits at the cap for most of the run)

Measured speed (generation tokens/sec)

ModelQuantEnginetok/sPrompt processing (tok/s)Details
Qwen3.6 35B-A3BNVFP4vLLM, b12x FP4 MoE kernel26126,600 @ 8,192-token prompt
est. TTFT @8k ≈ 0.31 s (calculated from measured prefill)
context: single stream📏 Pure decode probe, 512 forced output tokens, timed mid-generation: 3.8 ms per step (about 261 tok/s); 1,264 tok/s aggregate at batch 8, 3,769 at 64; NVFP4 beat two INT4 builds on the same card (3.9 vs 4.5 / 5.0 ms per token)

Prompt processing = prefill speed (how fast the prompt is read) as reported by the source, with the prompt length it was measured at; blank = the source does not report it. est. TTFT @8k is calculated from the measured prefill speed (8,192 ÷ prefill tok/s), shown only when prefill was measured on a prompt of 512+ tokens. It is not a measurement and ignores load time and caching. measured TTFT is the source's own time-to-first-token. Why prefill matters for coding assistants.

Owner-run test by Stefano Schotten (URE, 5 Nov 2026), used as the yardstick for a Mac Studio M5 Ultra (a separate RigBase build). End to end on a multi-turn RAG workload (about 11,300-token prompts): TPOT 3.9 ms and 196 tok/s output at 1 conversation, 836 tok/s at 32. CPU and RAM not stated.

Source: https://ure.us/articles/m5-ultra-vs-rtx-pro-6000/

Blank fields = not stated in the source. We never estimate build data; GPU street prices above are separately sourced and dated.

Embed this result

Copy into a README or blog. Links back to this build. See badge docs (CC BY 4.0 attribution). The SVG does not invent speeds — numbers stay on the build page.

Markdown

[![Measured on RigBase](https://getrigbase.com/badge/rtx-pro-6000-blackwell-max-q-qwen36-35b-a3b-nvfp4-vllm-ure.svg)](https://getrigbase.com/builds/rtx-pro-6000-blackwell-max-q-qwen36-35b-a3b-nvfp4-vllm-ure/)

HTML

<a href="https://getrigbase.com/builds/rtx-pro-6000-blackwell-max-q-qwen36-35b-a3b-nvfp4-vllm-ure/" rel="noopener"><img src="https://getrigbase.com/badge/rtx-pro-6000-blackwell-max-q-qwen36-35b-a3b-nvfp4-vllm-ure.svg" alt="Measured on RigBase" height="20"/></a>

Preview: Measured on RigBase · How to use

Rent before you buy: try this model on a cloud GPU for a few dollars first: Runpod (referral link; sign up with Google to get the new-user bonus)

Quick answers: Can RTX Pro 6000 Max-Q (Blackwell, 96 GB, 300 W) run Qwen3.6 35B-A3B?