NVIDIA
RTX PRO 6000 Blackwell Server Edition 96GB
RTX PRO BlackwellDatacenterBlackwellPCIe 5CUDA
Operating mode
Choose the operating mode for this hardware
Use this to bias workload recommendations toward responsiveness, background autonomy, lighter serving, or multi-GPU scale-out.
Current mode
Balanced
Balanced for general local use. Keeps the ranking neutral across personal and serving workflows.
About this GPU for AI
The RTX PRO 6000 Blackwell Server Edition is NVIDIA's most capable workstation-class accelerator, packing 96 GB of GDDR7 VRAM on Blackwell architecture for professional AI and visualization workloads. Its 120 TFLOPS of FP16 compute and 1,597 GB/s bandwidth make it suited for running 70B parameter models at Q4 with headroom to spare. As a PCIe 5.0 card, it slots into standard server platforms without the infrastructure requirements of SXM or NVLink systems. It bridges the gap between consumer workstations and full datacenter deployments.
Beyond LLMs
AI Capability Matrix
What AI tasks this GPU can handle โ from text generation to image and video creation.
| Capability | Status | Representative Model | Detail |
|---|
| LLM Chat (7B) | Runs natively | Llama 3.1 8B Q4 | โ |
| LLM Coding (30B) | Runs natively | Qwen 3 30B Q4 | โ |
| LLM Large (70B) |
massive-vramblackwell-architecturepcie-form-factorprofessional-grade
Specifications
Compute
FP16120 TFLOPS
INT84000 TOPS
ArchitectureBlackwell
Memory
VRAM96 GB
Bandwidth1597 GB/s
General
FamilyRTX PRO Blackwell
SegmentDatacenter
InterconnectPCIe 5
Compute PlatformCUDA
MSRP$9,999
Key Features
96 GB GDDR7 VRAM โ largest capacity in the RTX PRO lineupBlackwell architecture with 4th-gen Tensor Cores (FP4/FP8/FP16)1,597 GB/s memory bandwidth120 TFLOPS FP16 / 4,000 INT8 TOPSPCIe 5.0 x16 โ drop-in for modern server platforms300W TDP class โ single-slot power budget
For AI Workloads
Strengths
- 96 GB VRAM fits 70B models at Q4 and 34B models at FP16 on a single card
- Blackwell Tensor Cores with FP4 support deliver strong inference throughput per watt
- Standard PCIe 5.0 form factor works in any modern server โ no proprietary baseboard needed
- Commercially available at a fraction of H100/A100 pricing
Considerations
- GDDR7 bandwidth (1,597 GB/s) significantly below HBM-based datacenter GPUs like A100/H100
- No NVLink support limits multi-GPU scaling to PCIe peer-to-peer speeds
- Not suited for large-scale distributed training across GPU clusters
- ~$10K price point still steep for individual researchers or small teams
Blackwell is NVIDIA's fifth-generation RTX architecture, built on TSMC's 4NP process. It introduces 5th-generation Tensor Cores with native FP4 precision support, enabling double the inference throughput per watt compared to Ada Lovelace's FP8 operations. Key innovations include the Neural Rendering Pipeline for AI-driven shading and the debut of GDDR7 memory in consumer GPUs.
AI Relevance
FP4 Tensor Cores deliver the highest tokens-per-watt efficiency in any consumer architecture. Native FP4 quantization means models can run at lower precision with minimal quality loss, effectively doubling the effective VRAM for model weights.
Process: TSMC 4NPPlatform: CUDATensor Cores: Gen 5Precisions: FP32, FP16, BF16, FP8, FP4, INT8, INT4
Recommendations by Workload
Qwen 3.5 27B matches Chat and keeps a practical fit profile. It is a recent-generation family, which helps on current local SOTA workloads. It fits natively with comfortable headroom. Context coverage stays within the requested workload envelope. Known distribution channels: huggingface, ollama, lm-studio.
Decode 55.2 tok/s ยท 131K ctx ยท llama.cppEST.
Qwen3-Coder-Next is a specialized fit for Coding. It is a recent-generation family, which helps on current local SOTA workloads. It fits natively with comfortable headroom. Context coverage stays within the requested workload envelope. Known distribution channels: huggingface, ollama, lm-studio.
Decode 70.8 tok/s ยท 217K ctx ยท llama.cppEST.
Just out of reach
Models you could run with an upgrade
High-quality models that need a bit more memory
397BTier 100Needs ~254.1 GB
1000BTier 100Needs ~624.2 GB
1000BTier 100Needs ~624.2 GB
1600BTier 100Needs ~873.4 GB
284BTier 98Needs ~169.2 GB
Image & Video Generation
Diffusion Model Compatibility
51 of 52 models can generate images or video on your RTX PRO 6000 Blackwell Server Edition 96GB
Upgrade paths
Upgrade from RTX PRO 6000 Blackwell Server Edition 96GB
See what you unlock with more powerful hardware
Upgrade options
Upgrade options
141 GB VRAM (+45)4800 GB/s (+3203)
BUnlocks 2 additional models that do not fit on the current setup.Unlocks Qwen 3 235B A22B, MiniMax M2.7+55% faster avg
Unlocks 2 additional models that do not fit on the current setup.
Lifts average decode speed across fitting models by about 55%.
~$30,000 MSRP
180 GB VRAM (+84)8000 GB/s (+6403)
BUnlocks 8 additional models that do not fit on the current setup.Unlocks DeepSeek V4 Flash, Qwen 3 235B A22B, MiniMax M2.7+5 more ยท +89% faster avg
Unlocks 8 additional models that do not fit on the current setup.
Lifts average decode speed across fitting models by about 89%.
~$30,000 MSRP
AMD Instinct MI325X 256GBBiggest leap
256 GB VRAM (+160)6000 GB/s (+4403)
BUnlocks 12 additional models that do not fit on the current setup.Unlocks Qwen 3.5 397B A17B, DeepSeek V4 Flash, Qwen 3 235B A22B+9 more ยท +58% faster avg
Unlocks 12 additional models that do not fit on the current setup.
Lifts average decode speed across fitting models by about 58%.
~$20,000 MSRP
AMD Instinct MI350X 288GBBest value
288 GB VRAM (+192)8000 GB/s (+6403)
BUnlocks 13 additional models that do not fit on the current setup.Unlocks Qwen 3.5 397B A17B, DeepSeek V4 Flash, Qwen 3 235B A22B+10 more ยท +77% faster avg
Unlocks 13 additional models that do not fit on the current setup.
Lifts average decode speed across fitting models by about 77%.
~$8,000 MSRP
Frequently Asked Questions
RTX PRO 6000 Blackwell Server Edition 96GBCategory AvgNVIDIA H200 141GB
| Image Gen (SDXL) | Runs natively | SDXL 1.0 FP16 | ~~2.5s per image |
| Image Gen (Flux) | Runs natively | Flux.1 Dev FP16 | ~~11.1s per image |
| Image Gen (SD 3.5) | Runs natively | SD 3.5 Large FP16 | ~~13.6s per image |
| Video Short (25f) | Runs natively | LTX Video 2B | ~~2.1s/frame |
| Video Long (100f) | Tight fit | Wan Video 14B | ~~6.3s/frame |
S
Qwen3-Coder-Next is a specialized fit for Agentic Coding. It is a recent-generation family, which helps on current local SOTA workloads. It fits natively with comfortable headroom. Context coverage stays within the requested workload envelope. Known distribution channels: huggingface, ollama, lm-studio.
Decode 90.6 tok/s ยท 256K ctx ยท llama.cppEST.
Qwen3-Coder-Next matches Reasoning and keeps a practical fit profile. It is a recent-generation family, which helps on current local SOTA workloads. It fits natively with comfortable headroom. Context coverage stays within the requested workload envelope. Known distribution channels: huggingface, ollama, lm-studio.
Decode 70.8 tok/s ยท 217K ctx ยท llama.cppEST.
Qwen 3.5 27B matches RAG and keeps a practical fit profile. It is a recent-generation family, which helps on current local SOTA workloads. It fits natively with comfortable headroom. Context coverage stays within the requested workload envelope. Known distribution channels: huggingface, ollama, lm-studio.
Decode 55.2 tok/s ยท 131K ctx ยท llama.cppEST.
93
119B88.5 GB59 tok/s38K ctx
Image
| MAGI-1Video | 1280ร720 | ~5.8s/frame | S |
Image models estimated at 1024ร1024 (28 steps, FP16). Video models estimated at 768ร512 (25 frames, 30 steps, FP16). Actual performance varies with runtime and system load.
Buying advice
Should you buy RTX PRO 6000 Blackwell Server Edition 96GB for local AI?
Excellent choice for local AI
Runs 36 of 50 top models well โ a strong all-rounder for local inference.
What will limit you first
This setup is broadly balanced for this model.
No major red flags
This recommendation has enough memory headroom and acceptable estimated speed for the selected workload.
Best upgrade itinerary
Unlocks 2 additional models that do not fit on the current setup.
Want more headroom? NVIDIA H200 141GB (141.0 GB VRAM) is the next step up.