RTX PRO 6000 Blackwell
GPU only · GDDR7 · $8.5k launch, $15,499 on Amazon today
NVIDIAAdd-in GPU$15,499 · Aug 31, 2026View on Amazonpaid link
- Memory
- 96 GB~88 GB usable
- Bandwidth
- 1.8kGB/s — sets decode speed
- Compute
- 126FP16 TFLOPS — sets prefill
- Power
- 600 Wunder sustained load
- Noise
- 45 dBAyou will hear it
- Price per GB
- $161of accelerator memory
Price history
tracking
Tracking since 2026-08-31 — not enough history for a trend yet.
1 observation recorded · Last checked 2026-08-31. Check the retailer before acting on this — we would rather say the data is old than pretend otherwise.
What it can run
4 of 6 reference workloads fit. The largest is 120B MoE A5B · MXFP4 (gpt-oss class), at an estimated 260 tok/s of single-stream decode — the fastest of the 6 machines here that can run it.
| Workload | Needs | Fits | Prefill t/s | Decode t/s | Batched t/s | $ per t/s |
|---|---|---|---|---|---|---|
8B dense · Q4 (chat, drafts) dense · 4.7 GB read per token | 5.4 GB | Yes | 2.8k | 248 | 1290 | $62 |
32B dense · Q4 (coding) dense · 19 GB read per token | 21.8 GB | Yes | 689 | 61 | 317 | $254 |
70B dense · Q4 (strong general) dense · 40 GB read per token | 46 GB | Yes | 315 | 29 | 151 | $534 |
120B MoE A5B · MXFP4 (gpt-oss class) MoE · 2.9 GB read per token | 72.4 GB | Yes | 4.3k | 260 | 1352 | $60 |
320B MoE A18B · FP8 (frontier-ish) MoE · 18 GB read per token | 368 GB | No | 1.2k | 42 | 218 | — |
671B MoE A37B · Q4 (DeepSeek class) MoE · 18.5 GB read per token | 437 GB | No | 596 | 41 | 213 | — |
Figures are modelled from this machine's bandwidth and compute, not measured on one. How they are derived →
Alternatives that also run 120B MoE A5B · MXFP4 (gpt-oss class)
| Machine | Price | Memory | Decode t/s | $ per t/s |
|---|---|---|---|---|
| $4,500 | 128 GB | 89 | $51 | |
RTX PRO 6000 Blackwell — this machine | $15,499 | 96 GB | 260 | $60 |
| $13,500 | 512 GB | 174 | $78 | |
| $9,499 | 512 GB | 119 | $80 | |
| $3,999 | 128 GB | 40 | $100 | |
| $5,199 | 128 GB | 40 | $130 |