GPU Specifications / NVIDIA
NVIDIA RTX PRO 6000 Blackwell
Blackwell datacenter GPU with 96 GB of GDDR7 memory, 1792 GB/s of bandwidth and up to 1,000 TFLOPS of FP16 tensor compute.
Key Specifications
Memory
96 GB GDDR7
Memory Bandwidth
1,792 GB/s
TDP
600 W
Architecture
Blackwell
Interconnect
PCIe 5.0 · 128 GB/s
Est. On-demand Price
~$5.00/h
Hourly rates are indicative on-demand estimates; actual pricing varies by provider and commitment.
Compute Performance
| Precision | Peak throughput |
|---|---|
| FP64 (double precision) | 2 TFLOPS |
| FP32 (single precision) | 125 TFLOPS |
| FP24 | 250 TFLOPS |
| FP16 (tensor) | 1,000 TFLOPS |
| INT8 (tensor) | 2,000 TFLOPS |
| INT4 (tensor) | 4,000 TFLOPS |
Tensor figures use the vendor's peak numbers (with structured sparsity where supported).
System Requirements
Recommended CPU
AMD EPYC 9554 or Intel Xeon Platinum 8580
Max VRAM per node (8 GPUs)
768 GB
System RAM (min / recommended)
256 / 512 GB
Minimum PSU
1200 W
LLMs on the RTX PRO 6000 Blackwell
Number of RTX PRO 6000 Blackwell GPUs needed to serve popular models at 8-bit quantization (including 20% overhead for activations and KV cache).
| Model | Params | VRAM (8-bit) | GPUs needed |
|---|---|---|---|
| GPT-5.6 Sol | 2400B | 2682 GB | 28x RTX PRO 6000 Blackwell |
| GPT-5 Flagship | 2100B | 2347 GB | 25x RTX PRO 6000 Blackwell |
| GPT-5.6 Luna | 1400B | 1565 GB | 17x RTX PRO 6000 Blackwell |
| Kimi K3 (1.2T) | 1200B | 1341 GB | 14x RTX PRO 6000 Blackwell |
| Kimi K2.6 (1T) | 1000B | 1118 GB | 12x RTX PRO 6000 Blackwell |
| GPT-5.6 Terra | 800B | 894 GB | 10x RTX PRO 6000 Blackwell |
| DeepSeek V4 Pro (671B) | 671B | 750 GB | 8x RTX PRO 6000 Blackwell |
| Llama 4 Behemoth (500B) | 500B | 559 GB | 6x RTX PRO 6000 Blackwell |
| Claude 5 Fable (480B) | 480B | 536 GB | 6x RTX PRO 6000 Blackwell |
| GLM 5.2 (400B) | 400B | 447 GB | 5x RTX PRO 6000 Blackwell |
| Grok 4.5 | 350B | 391 GB | 5x RTX PRO 6000 Blackwell |
| Claude 4.8 Opus (300B) | 300B | 335 GB | 4x RTX PRO 6000 Blackwell |
| Muse Spark 1.1 | 300B | 335 GB | 4x RTX PRO 6000 Blackwell |
| Grok 4 | 270B | 302 GB | 4x RTX PRO 6000 Blackwell |
| Gemini 3.1 Pro | 250B | 279 GB | 3x RTX PRO 6000 Blackwell |
| Qwen 3.7 Max (235B) | 235B | 263 GB | 3x RTX PRO 6000 Blackwell |
| Mistral Large 3 (200B) | 200B | 224 GB | 3x RTX PRO 6000 Blackwell |
| Grok 3 Mini | 190B | 212 GB | 3x RTX PRO 6000 Blackwell |
| Claude 5 Sonnet (175B) | 175B | 196 GB | 3x RTX PRO 6000 Blackwell |
| Gemini 3.5 Flash | 150B | 168 GB | 2x RTX PRO 6000 Blackwell |
| Gemini 2.5 Flash | 140B | 156 GB | 2x RTX PRO 6000 Blackwell |
| Llama 4 Maverick (128B) | 128B | 143 GB | 2x RTX PRO 6000 Blackwell |
| DeepSeek V4 Flash (120B) | 120B | 134 GB | 2x RTX PRO 6000 Blackwell |
| Qwen 3.6 Plus (110B) | 110B | 123 GB | 2x RTX PRO 6000 Blackwell |
| Nova Premier (80B) | 80B | 89 GB | 1x RTX PRO 6000 Blackwell |
| Qwen 3 Coder-Next (80B) | 80B | 89 GB | 1x RTX PRO 6000 Blackwell |
| Claude 4.5 Haiku (70B) | 70B | 78 GB | 1x RTX PRO 6000 Blackwell |
| Llama 3.3 Instruct (70B) | 70B | 78 GB | 1x RTX PRO 6000 Blackwell |
| Mistral Medium 3.5 (70B) | 70B | 78 GB | 1x RTX PRO 6000 Blackwell |
| Yi 1.5 (40B) | 40B | 45 GB | 1x RTX PRO 6000 Blackwell |
| Nova Core (34B) | 34B | 38 GB | 1x RTX PRO 6000 Blackwell |
| DeepSeek V3.1 (32B) | 32B | 36 GB | 1x RTX PRO 6000 Blackwell |
| Gemma 3 (27B) | 27B | 30 GB | 1x RTX PRO 6000 Blackwell |
| Mistral Small 4 (24B) | 24B | 27 GB | 1x RTX PRO 6000 Blackwell |
| Yi 1.5 (15B) | 15B | 17 GB | 1x RTX PRO 6000 Blackwell |
| Phi 4 (14B) | 14B | 16 GB | 1x RTX PRO 6000 Blackwell |
| Nova Lite (12B) | 12B | 13 GB | 1x RTX PRO 6000 Blackwell |
| Llama 3.2 Instruct (11B) | 11B | 12 GB | 1x RTX PRO 6000 Blackwell |
| Gemma 3 (9B) | 9B | 10 GB | 1x RTX PRO 6000 Blackwell |
| Yi 1.5 Lite (9B) | 9B | 10 GB | 1x RTX PRO 6000 Blackwell |
| Phi 4 Mini (7B) | 7B | 8 GB | 1x RTX PRO 6000 Blackwell |
| Phi 3.5 (3.8B) | 3.8B | 4 GB | 1x RTX PRO 6000 Blackwell |
Frequently Asked Questions
How much VRAM does the NVIDIA RTX PRO 6000 Blackwell have?
The NVIDIA RTX PRO 6000 Blackwell has 96 GB of GDDR7 memory with 1792 GB/s of memory bandwidth.
Which LLMs can run on a single RTX PRO 6000 Blackwell?
At 8-bit quantization, a single RTX PRO 6000 Blackwell (96 GB) can serve models up to roughly 80B parameters, such as Nova Premier (80B). Larger models require multiple GPUs or more aggressive quantization.
How much does it cost to rent a NVIDIA RTX PRO 6000 Blackwell?
On-demand cloud pricing for the RTX PRO 6000 Blackwell is around $5.00/hour, i.e. about $3,650/month running 24/7. Actual prices vary by provider, region, and commitment.
What are the power and system requirements of the NVIDIA RTX PRO 6000 Blackwell?
The RTX PRO 6000 Blackwell has a TDP of 600W. A power supply of at least 1200W per GPU is recommended. Recommended host CPUs: AMD EPYC 9554 or Intel Xeon Platinum 8580.
Deploy on a GPU cloud
Rent the NVIDIA RTX PRO 6000 Blackwell by the hour instead of buying hardware.
