NVIDIA · Blackwell
Best AI Models for NVIDIA RTX PRO 6000 Blackwell Server Edition (96 GB)
Runs everything up to 120B MoE well.
96 GB GDDR71597 GB/s500 TFLOPS FP16600 WBlackwell 2025
Passive 96 GB server variant — needs chassis airflow; commonly rented per-GPU rather than bought for a desktop.
| # | Model | Speed | llmrun Score | Quant · VRAM |
|---|---|---|---|---|
| 1 | Qwen3.8 27BAlibabaHow to run → Q4_K_M · 17.9 GB · 59.6 t/s | 59.6 tok/s | Q4_K_M · 17.9 GB | |
| 2 | Qwen3.6 27BAlibaba Q4_K_M · 17.9 GB · 59.6 t/s | 59.6 tok/s | Q4_K_M · 17.9 GB | |
| 3 | Gemma 4 31B ITGoogle Q4_K_M · 21.7 GB · 50.9 t/s | 50.9 tok/s | Q4_K_M · 21.7 GB | |
| 4 | Gemma 4 26B A4B ITGoogle Q4_K_M · 16.5 GB · 233 t/s | 233 tok/s | Q4_K_M · 16.5 GB | |
| 5 | Q4_K_M · 19.3 GB · 235 t/s | 235 tok/s | Q4_K_M · 19.3 GB |
| # | Model | Speed | Coding Score | Quant · VRAM |
|---|---|---|---|---|
| 1 | Qwen3.8 27BAlibabaHow to run → Q4_K_M · 17.9 GB · 59.6 t/s | 59.6 tok/s | Q4_K_M · 17.9 GB | |
| 2 | Gemma 4 31B ITGoogle Q4_K_M · 21.7 GB · 50.9 t/s | 50.9 tok/s | Q4_K_M · 21.7 GB | |
| 3 | Qwen3.6 35B A3BAlibaba Q4_K_M · 22.0 GB · 200 t/s | 200 tok/s | Q4_K_M · 22.0 GB | |
| 4 | Gemma 4 26B A4B ITGoogle Q4_K_M · 16.5 GB · 233 t/s | 233 tok/s | Q4_K_M · 16.5 GB | |
| 5 | Q4_K_M · 19.3 GB · 235 t/s | 235 tok/s | Q4_K_M · 19.3 GB |
| # | Model | Speed | Reasoning Score | Quant · VRAM |
|---|---|---|---|---|
| 1 | Qwen3.8 27BAlibabaHow to run → Q4_K_M · 17.9 GB · 59.6 t/s | 59.6 tok/s | Q4_K_M · 17.9 GB | |
| 2 | Gemma 4 31B ITGoogle Q4_K_M · 21.7 GB · 50.9 t/s | 50.9 tok/s | Q4_K_M · 21.7 GB | |
| 3 | Qwen3.5 27BAlibaba Q4_K_M · 17.9 GB · 59.6 t/s | 59.6 tok/s | Q4_K_M · 17.9 GB | |
| 4 | Gemma 4 26B A4B ITGoogle Q4_K_M · 16.5 GB · 233 t/s | 233 tok/s | Q4_K_M · 16.5 GB | |
| 5 | Q4_K_M · 19.3 GB · 235 t/s | 235 tok/s | Q4_K_M · 19.3 GB |
| # | Model | Speed | Math Score | Quant · VRAM |
|---|---|---|---|---|
| 1 | Qwen3.8 27BAlibabaHow to run → Q4_K_M · 17.9 GB · 59.6 t/s | 59.6 tok/s | Q4_K_M · 17.9 GB | |
| 2 | Qwen3.6 27BAlibaba Q4_K_M · 17.9 GB · 59.6 t/s | 59.6 tok/s | Q4_K_M · 17.9 GB | |
| 3 | Qwen3.6 35B A3BAlibaba Q4_K_M · 22.0 GB · 200 t/s | 200 tok/s | Q4_K_M · 22.0 GB | |
| 4 | Gemma 4 31B ITGoogle Q4_K_M · 21.7 GB · 50.9 t/s | 50.9 tok/s | overall | Q4_K_M · 21.7 GB |
| 5 | Q4_K_M · 19.3 GB · 235 t/s | 235 tok/s | Q4_K_M · 19.3 GB |
| # | Model | Speed ↓ | llmrun Score | Quant · VRAM |
|---|---|---|---|---|
| 1 | Qwen3.5 4BAlibabaHow to run → Q4_K_M · 3.4 GB · 318 t/s | 318 tok/s | Q4_K_M · 3.4 GB | |
| 2 | GPT OSS 20BOpenAI Q4_K_M · 13.1 GB · 262 t/s | 262 tok/s | Q4_K_M · 13.1 GB | |
| 3 | Q4_K_M · 19.3 GB · 235 t/s | 235 tok/s | Q4_K_M · 19.3 GB | |
| 4 | Gemma 4 26B A4B ITGoogle Q4_K_M · 16.5 GB · 233 t/s | 233 tok/s | Q4_K_M · 16.5 GB | |
| 5 | Qwen3 30B A3B Thinking 2507Alibaba Q4_K_M · 19.4 GB · 216 t/s | 216 tok/s | Q4_K_M · 19.4 GB |
| # | Model | Speed | Vision Score | Quant · VRAM |
|---|---|---|---|---|
| 1 | Qwen3.8 27BAlibabaHow to run → Q4_K_M · 17.9 GB · 59.6 t/s | 59.6 tok/s | Q4_K_M · 17.9 GB | |
| 2 | Qwen3.6 27BAlibaba Q4_K_M · 17.9 GB · 59.6 t/s | 59.6 tok/s | Q4_K_M · 17.9 GB | |
| 3 | Gemma 4 31B ITGoogle Q4_K_M · 21.7 GB · 50.9 t/s | 50.9 tok/s | Q4_K_M · 21.7 GB | |
| 4 | Gemma 4 26B A4B ITGoogle Q4_K_M · 16.5 GB · 233 t/s | 233 tok/s | Q4_K_M · 16.5 GB | |
| 5 | JEV 27B VLautotrust Q4_K_M · 17.9 GB · 59.6 t/s | 59.6 tok/s | Q4_K_M · 17.9 GB |
See all 391 models on the RTX PRO 6000 Blackwell Server Edition →Best quant that fits · speeds estimated on this card
The RTX PRO 6000 Blackwell Server Edition runs everything up to 120B MoE well. Blackwell generation, released 2025.
GenerationBlackwellreleased 2025
RTX PRO 6000 Blackwell Server Edition strengths and weaknesses
Strengths
- Runs 120B MoE models: GPT OSS 120B at 188 tok/s in 70.8 GB
- Room for long context: Llama 3.3 70B Instruct at Q4 uses 51.3 GB, leaving 44.7 GB free
- Reads long prompts fast: A 4K-token prompt on Qwen3.8 27B starts answering in 1.0 s
- FP8 and NVFP4 run natively: Blackwell takes FP8 and NVFP4 checkpoints in vLLM as published
Weaknesses
- 600 W under load: Plan a 1,200 W PSU; ≈$131 a year at 4 h a day
Output speed depends on memory bandwidth (1597 GB/s).
- 9BQwen3.5 9B · Q4_K_M · 6.6 GB163tok/s
- 14BQwen3 14B · Q4_K_M · 11.9 GB109tok/s
- 27–31BQwen3.8 27B · Q4_K_M · 17.9 GB59.6tok/s
- 35B MoEQwen3.6 35B A3B · Q4_K_M · 22.0 GB200tok/s
- 70BLlama 3.3 70B Instruct · Q4_K_M · 51.3 GB22.3tok/s
- 120B MoEGPT OSS 120B · Q4_K_M · 70.8 GB188tok/s
Estimated speeds: Q4_K_M, 4K context, one user. How we estimate
All 1270 models on the RTX PRO 6000 Blackwell Server Edition
Featured onlyBest quant that fits, ordered by llmrun Score. Fit grades how well it runs here, S best. How ranking works
| Model | Quant | VRAM | Speed | Context | Score | Fit |
|---|---|---|---|---|---|---|
NuExtract 1.5Chat | Q4_K_M | 9.0 GB | 305.3 t/s | 131K | S100, Runs great | |
Tiny MixtralChat | Q4_K_M | 0.5 GB | 3209.4 t/s | 131K | S100, Runs great | |
| Q4_K_M | 1.7 GB | 1193.2 t/s | 33K | S100, Runs great | ||
| Q4_K_M | 7.7 GB | 160.7 t/s | 8K | S97, Runs great | ||
Supra 50M BaseChat | Q4_K_M | 0.3 GB | 2883.5 t/s | 1K | S100, Runs great | |
Granite 4.0 MicroChatFunctions | Q4_K_M | 3.7 GB | 413.6 t/s | 131K | S100, Runs great | |
LFM2.5 1.2B InstructChatFunctions | Q4_K_M | 1.2 GB | 970.1 t/s | 128K | S100, Runs great | |
Gemma 4 12BChat | Q4_K_M | 9.0 GB | 126.1 t/s | 262K | S95, Runs great | |
Distilgpt2Chat | Q4_K_M | 0.1 GB | 17300.8 t/s | 1K | S100, Runs great | |
Gemma 4 E2BChat | Q4_K_M | 3.5 GB | 302.6 t/s | 131K | S100, Runs great | |
| Q4_K_M | 6.2 GB | 192.6 t/s | 8K | S99, Runs great | ||
Aya Expanse 8BChat | BF16 | 20.2 GB | 58.8 t/s | — | A83, Runs well | |
Granite 3.3 2B InstructChatFunctions | Q4_K_M | 3.2 GB | 521.6 t/s | 131K | S100, Runs great | |
Llama 3.2 11B Vision InstructVisionFunctions | Q4_K_M | 9.8 GB | 147.5 t/s | — | S97, Runs great | |
Llama 3.3 70B Instruct AbliteratedChatFunctions | Q4_K_M | 48 GB | 24.0 t/s | 131K | B64, Decent |
Rent NVIDIA RTX PRO 6000 Blackwell Server Edition in the cloud
Hourly rate on demand for one RTX PRO 6000 Blackwell Server Edition.
Cost per million tokens
Output-token cost on an RTX PRO 6000 Blackwell Server Edition rented at $1.55/hr, one request at a time. Batch several users on the same card and the cost per token drops several-fold.
| Model | Quant | Speed | Per 1M output |
|---|---|---|---|
| NVIDIA Nemotron 3.5 Lightning 30B A3B BF16 | 235 tok/s | $1.83$0.0090 input | |
| Gemma 4 26B A4B IT | 233 tok/s | $1.85$0.017 input | |
| Qwen3.8 27B | 60 tok/s | $7.22$0.11 input | |
| Muse Glimmer 30B | 57 tok/s | $7.60$0.11 input | |
| Gemma 4 31B IT | 51 tok/s | $8.45$0.12 input |
Specs
- Architecture
- Blackwell
- VRAM
- 96 GB GDDR7
- Memory bandwidth
- 1597.0 GB/s
- FP16 (dense)
- 500.00 TFLOPS
- CUDA cores
- 24,064
- Tensor cores
- 752
- Compute capability
- 12.0
- TDP
- 600 W
- Released
- 2025-05-01
RTX PRO 6000 Blackwell Server Edition FAQ
What is the largest model the RTX PRO 6000 Blackwell Server Edition can run?
Mistral Medium 3.5 128B (127.7B) is the largest at 4-bit or better: at Q4_K_M it uses 82.8 GB of 96 GB, and runs at about 13.4 tok/s.
Can the RTX PRO 6000 Blackwell Server Edition run Qwen3.8 27B?
Yes. Qwen3.8 27B fits at Q4_K_M, using 17.9 GB of 96 GB, and it is the best model that runs well on this card.
How many tokens per second does the RTX PRO 6000 Blackwell Server Edition give?
About 59.6 tok/s on Qwen3.8 27B at Q4_K_M, with the first token after 1.0 s on a 4K-token prompt.
What power supply does the RTX PRO 6000 Blackwell Server Edition need?
Plan a 1200 W PSU (2× the 600 W TDP).
Is 96 GB enough for coding and agents?
Yes for coding: Qwen3.8 27B, the best of the 13 coding models that fit, runs at about 60 tok/s.