NVIDIA RTX Pro 6000
The NVIDIA RTX Pro 6000 has 96 GB VRAM and 1792 GB/s memory bandwidth. It can run 69 of our 99 tracked models natively in VRAM at 8k context.
With 96 GB GDDR7, the NVIDIA RTX Pro 6000 is a workstation-tier GPU that can run 69 models natively.
The NVIDIA RTX Pro 6000 (Blackwell Workstation Edition) is the flagship Blackwell workstation GPU, doubling the RTX 6000 Ada's VRAM to 96GB of ECC GDDR7 on a 512-bit bus at 1,792 GB/s, the same bandwidth as the RTX 5090. It uses a nearly full GB202 die with 24,064 CUDA cores (the same silicon as the RTX 5090's cut-down 21,760) in a 600W dual-slot workstation card with professional drivers and error-correcting memory. NVIDIA's datasheet lists no NVLink, so a second card pools over PCIe 5.0 x16 only. The 300W Max-Q variant carries the same memory and bandwidth at lower clocks, while the passively cooled Server Edition is rated at up to 1,597 GB/s; this entry tracks the Workstation Edition. The 96GB capacity is large enough to run 70B models at Q4_K_M or Q8_0 entirely in VRAM without any CPU offloading, and comfortably holds multiple models simultaneously. At roughly $8,565 at launch (NVIDIA publishes no conventional MSRP, and its own marketplace price has since risen), it is the definitive single-GPU option for on-prem LLM inference when model fit and professional reliability matter more than cost.
NVIDIA RTX Pro 6000: 2025 Blackwell flagship workstation GPU built on a nearly full GB202 die (24,064 CUDA cores, the same silicon as the RTX 5090) with 96GB of ECC GDDR7 on a 512-bit bus at 1,792 GB/s, double the RTX 6000 Ada's capacity, at roughly $8,565 at launch.
70B models fit entirely in VRAM at Q4_K_M or even Q8_0 with no CPU offload, and it comfortably holds multiple smaller models loaded simultaneously. ~35-50 t/s for 7B Q4.
Full CUDA with ECC and PCIe 5.0 x16, but no NVLink (NVIDIA's datasheet lists none), so multi-card systems pool over PCIe. The 600W Workstation Edition is poorly suited to dense multi-GPU builds; the 300W Max-Q variant exists for that. The definitive single-GPU option for on-prem LLM inference when model fit matters more than cost; no consumer card at any price offers this much VRAM in one slot.
| Vendor | NVIDIA |
| Architecture | Blackwell |
| VRAM | 96 GB |
| Memory type | GDDR7 |
| Memory bandwidth | 1792 GB/s |
| Compute backend | CUDA |
| Tier | Workstation |
| Released | 2025 |
| Models (native) | 69 / 99 |
| Models (offload) | 4 / 99 |
Cloud GPU Rental
Don't want to buy an NVIDIA RTX Pro 6000? RunPod is a cloud GPU rental service: rent one by the hour instead, no contract, no upfront hardware cost.
Pay by the hour · no contract · pods start in about a minute.
Rent an RTX Pro 6000 from $1.69/hr on RunPod ↗ (+$5 signup credit)Prices as of Oct 13, 2026. Rates change; check RunPod for current pricing.
Affiliate link: CanItRun may earn a commission. Doesn't affect the fit calculation above.
Popular models for this GPU
Models this GPU runs natively in VRAM (69)
- Step 3.7 Flash198B · MMLU-Pro N/AQ2_K · ~136.9 t/s
- Step 3.5 Flash196.81B · MMLU-Pro 84.4Q2_K · ~136.9 t/s
- Qwen3.8-Flash-Next180B · MMLU-Pro N/AQ2_K · ~181.4 t/s
- Mixtral 8x22B Instruct v0.1141B · MMLU-Pro 40.0NVFP4 · ~55.1 t/s
- Mistral Medium 3.5 128B128B · MMLU-Pro N/ANVFP4 · ~20.9 t/s
Show 64 more
- Qwen 3.5 122B-A10B (MoE)122B · MMLU-Pro 86.7NVFP4 · ~138.2 t/s
- Nemotron 3 Super 120B120B · MMLU-Pro 83.7NVFP4 · ~87 t/s
- GPT-OSS 120B117B · MMLU-Pro 80.7NVFP4 · ~205 t/s
- Llama 4 Scout 109B109B · MMLU-Pro 74.3NVFP4 · ~73.2 t/s
- GLM-4.5 Air 106B106B · MMLU-Pro 81.4NVFP4 · ~116.9 t/s
- GLM-4.6V 106B106B · MMLU-Pro 79.9NVFP4 · ~116.9 t/s
- Qwen 2.5 72B Instruct72B · MMLU-Pro 71.1NVFP4 · ~34.1 t/s
- Llama 3.3 70B Instruct70B · MMLU-Pro 68.9NVFP4 · ~34.8 t/s
- DeepSeek R1 Distill Llama 70B70B · MMLU-Pro 70.0NVFP4 · ~34.8 t/s
- Llama 3.1 70B Instruct70B · MMLU-Pro 66.4NVFP4 · ~34.8 t/s
- Mixtral 8x7B Instruct v0.146.7B · MMLU-Pro 29.7NVFP4 · ~135.2 t/s
- Command-R 35B35B · MMLU-Pro 33.0BF16 · ~18.7 t/s
- Qwen 3.5 35B-A3B (MoE)35B · MMLU-Pro 85.3BF16 · ~139.1 t/s
- Qwen 3.6 35B-A3B (MoE)35B · MMLU-Pro 85.2BF16 · ~139.1 t/s
- Ornith 1.5 35B-A3B (MoE)35B · MMLU-Pro N/ABF16 · ~139.1 t/s
- Yi 1.5 34B Chat34.4B · MMLU-Pro 37.0BF16 · ~20.6 t/s
- Qwen3 32B32.8B · MMLU-Pro 65.5BF16 · ~21.6 t/s
- Qwen 2.5 32B Instruct32.5B · MMLU-Pro 69.0BF16 · ~21.5 t/s
- Qwen 2.5 Coder 32B Instruct32.5B · MMLU-Pro 62.3BF16 · ~21.5 t/s
- DeepSeek R1 Distill Qwen 32B32.5B · MMLU-Pro 65.0BF16 · ~21.5 t/s
- Nemotron 3 Nano 30B32B · MMLU-Pro 78.3BF16 · ~123.4 t/s
- Gemma 4 31B30.7B · MMLU-Pro 85.2BF16 · ~23 t/s
- Qwen3 30B-A3B (MoE)30B · MMLU-Pro 61.5BF16 · ~121.2 t/s
- Nemotron 3.5 Lightning 30B-A3B30B · MMLU-Pro 81.6BF16 · ~118 t/s
- Muse Glimmer 30B27.8B · MMLU-Pro N/ABF16 · ~26 t/s
- Gemma 2 27B Instruct27.2B · MMLU-Pro 38.0BF16 · ~25.5 t/s
- Gemma 3 27B Instruct27B · MMLU-Pro 67.5BF16 · ~26.3 t/s
- Qwen 3.6 27B27B · MMLU-Pro 86.2BF16 · ~26 t/s
- UI-Mate 27B27B · MMLU-Pro ~86.2BF16 · ~26 t/s
- Bonsai 27B27B · MMLU-Pro ~81.5Ternary (Q2_0) · ~114.3 t/s
- Bonsai 2 27B27B · MMLU-Pro N/ATernary (Q2_0) · ~126.2 t/s
- Qwen 3.8 27B27B · MMLU-Pro N/ABF16 · ~26 t/s
- Gemma 4 26B (MoE)25.2B · MMLU-Pro 82.6BF16 · ~133 t/s
- Mistral Small 3.1 24B Instruct24B · MMLU-Pro 66.8BF16 · ~29.7 t/s
- Mistral Small 22B22.2B · MMLU-Pro 49.2BF16 · ~30.6 t/s
- GPT-OSS 20B21B · MMLU-Pro 67.9BF16 · ~150.1 t/s
- Qwen3 14B14.8B · MMLU-Pro 61.0BF16 · ~45.4 t/s
- Qwen 2.5 14B Instruct14.7B · MMLU-Pro 63.7BF16 · ~44.4 t/s
- Phi-4 14B Instruct14B · MMLU-Pro 70.4BF16 · ~47.1 t/s
- Mistral Nemo 12B Instruct12.2B · MMLU-Pro 35.6BF16 · ~53.4 t/s
- Gemma 3 12B Instruct12.2B · MMLU-Pro 60.6BF16 · ~53.6 t/s
- Gemma 4 12B (Unified)12B · MMLU-Pro 77.2BF16 · ~54.4 t/s
- Gemma 2 9B Instruct9.2B · MMLU-Pro 32.0BF16 · ~62.6 t/s
- Qwen 3.5 9B9B · MMLU-Pro 82.5BF16 · ~74 t/s
- Ornith 1.5 9B9B · MMLU-Pro N/ABF16 · ~74 t/s
- Llama 3.1 8B Instruct8B · MMLU-Pro 48.3BF16 · ~78.4 t/s
- DeepSeek R1 Distill Llama 8B8B · MMLU-Pro 41.0BF16 · ~78.4 t/s
- Qwen3 8B8B · MMLU-Pro 56.7BF16 · ~76.5 t/s
- Qwen 2.5 7B Instruct7.6B · MMLU-Pro 56.3BF16 · ~86.1 t/s
- Gemma 4 E4B7.52B · MMLU-Pro 69.4BF16 · ~120.1 t/s
- Mistral 7B Instruct v0.37.25B · MMLU-Pro 30.0BF16 · ~84.8 t/s
- Gemma 4 E2B4.65B · MMLU-Pro 60.0BF16 · ~197.9 t/s
- Gemma 3 4B Instruct4B · MMLU-Pro 43.6BF16 · ~139 t/s
- Phi-3.5 Mini Instruct3.8B · MMLU-Pro 47.4BF16 · ~113.9 t/s
- Phi-4-mini Instruct3.8B · MMLU-Pro 67.3BF16 · ~134.8 t/s
- Llama 3.2 3B Instruct3.2B · MMLU-Pro 24.0BF16 · ~157.9 t/s
- Qwen 2.5 3B Instruct3.1B · MMLU-Pro 32.4BF16 · ~159.2 t/s
- Gemma 2 2B Instruct2.6B · MMLU-Pro 17.8BF16 · ~184.8 t/s
- SmolLM2 1.7B Instruct1.7B · MMLU-Pro 19.0BF16 · ~216.6 t/s
- Qwen 2.5 1.5B Instruct1.5B · MMLU-Pro 16.8BF16 · ~268 t/s
- Llama 3.2 1B Instruct1.24B · MMLU-Pro 12.5BF16 · ~370 t/s
- Gemma 3 1B Instruct1B · MMLU-Pro 14.7BF16 · ~329.4 t/s
- Qwen 2.5 0.5B Instruct0.5B · MMLU-Pro 10.0BF16 · ~467.7 t/s
- SmolLM2 360M Instruct0.36B · MMLU-Pro 8.0BF16 · ~386.2 t/s
Models that fit with CPU offload (4)
These use system RAM for layers that don't fit in VRAM, so expect much slower inference.
Too large for this GPU (26)
- Llama 3.1 405B Instruct
- DeepSeek V3 671B
- DeepSeek R1 671B
- Llama 4 Maverick 400B
- MiniMax M1 456B
- GLM-4.5 355B
- GLM-4.6 355B
- GLM-4.7 358B
- GLM-5 744B
- Kimi K2.6
- GLM-5.1 754B
- DeepSeek V4 Pro 1.6T
- DeepSeek V4 Flash 284B
- GLM-5.2 753B
- Nemotron 3 Ultra 550B-A55B
- MiMo V2.5 Pro
- Kimi K2.5
- MiniMax M3
- Inkling
- Kimi K3
- Qwen3.8 2.4T-A95B
- DeepSeek V4 Pro 0813 1.6T
- Ornith 1.5 397B (MoE)
- GLM-5.3 753B
- GLM-5.3-Flash 320B
- DeepSeek V4.1 Flash 552B
Compare NVIDIA RTX Pro 6000 with other GPUs
- NVIDIA RTX Pro 6000vsNVIDIA RTX 6000 Ada+48 GB VRAM
- NVIDIA RTX Pro 6000vsNVIDIA RTX 5090+64 GB VRAM
- NVIDIA RTX Pro 6000vsNVIDIA L40S+48 GB VRAM
- NVIDIA RTX Pro 6000vsNVIDIA DGX Spark (128GB)-32 GB VRAM
- NVIDIA RTX Pro 6000vsAMD Radeon AI PRO R9700 32GB+64 GB VRAM
- NVIDIA RTX Pro 6000vsApple M3 Ultra (96GB)96 GB each
- NVIDIA RTX Pro 6000vsNVIDIA H100 80GB+16 GB VRAM
Frequently asked questions
- How much VRAM does the NVIDIA RTX Pro 6000 have?
- The NVIDIA RTX Pro 6000 has 96 GB of GDDR7 with 1792 GB/s memory bandwidth.
- What is the NVIDIA RTX Pro 6000 best for?
- With 96 GB of VRAM, the NVIDIA RTX Pro 6000 is a high-capacity workstation platform that runs 70B-class dense models and large MoE models natively, with plenty of room for long context.
- What LLMs can the NVIDIA RTX Pro 6000 run locally?
- The NVIDIA RTX Pro 6000 can run 69 of the 99 open-weight models tracked by CanItRun natively in VRAM at 8k context. Top options include: Qwen3.8-Flash-Next at Q2_K, Qwen 3.8 27B at BF16, Ornith 1.5 35B-A3B (MoE) at BF16.
- Can the NVIDIA RTX Pro 6000 run Gemma 4 31B?
- Yes. The NVIDIA RTX Pro 6000 runs Gemma 4 31B natively in VRAM at BF16 quantization, achieving approximately 23 tokens per second.
- Can the NVIDIA RTX Pro 6000 run Qwen 3.6 27B?
- Yes. The NVIDIA RTX Pro 6000 runs Qwen 3.6 27B natively in VRAM at BF16 quantization, achieving approximately 26 tokens per second.
- Can the NVIDIA RTX Pro 6000 run Qwen3 8B?
- Yes. The NVIDIA RTX Pro 6000 runs Qwen3 8B natively in VRAM at BF16 quantization, achieving approximately 76.5 tokens per second.
- Can I rent the NVIDIA RTX Pro 6000 instead of buying it?
- Yes: RunPod and similar cloud GPU providers let you rent NVIDIA RTX Pro 6000 instances by the hour, with no long-term contract. Rentals start from $1.69/hr on RunPod (as of Oct 13, 2026). This is often cheaper than buying if you only need it occasionally, and lets you try the GPU before committing to a purchase.