- Services
Every open model in our directory, ranked by how well it runs on NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition (96GB) at 4-bit. Grades weigh whether the model fits in memory and how fast it should run.
NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition has 96 GB of VRAM, enough to run 280 of the 336 open models we track at 4-bit. The largest that still fits well is AliceAI-Foundation-80B-A3B-Base (81.3B), and the top-rated is Kolibri at about 55 tok/s.
280 of 336 models run comfortably on NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition.
See full NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition specs and pricingThe top models this device can run at 4-bit, ranked by fit and speed.
| Model | Grade | Speed | VRAM |
|---|---|---|---|
| KolibriAleph Alpha | SS | 55.2 tok/s | 69.7 GB |
| AliceAI-Foundation-80B-A3B-Baseyandex | SS | 169.8 tok/s | 53.9 GB |
| Nemotron 3.5 Bolt 74B-A9B 24GBNVIDIA | SS | 240.9 tok/s | 43.7 GB |
| NVIDIA Nemotron 3.5 Bolt 74B-A9B 24GBNVIDIA | SS | 240.9 tok/s | 43.7 GB |
| Comet 60BMultiverse Computing | SS | 173.7 tok/s | 36.4 GB |
| K2-Horizon-MoVA-36B-A4BInstitute of Foundation Models | SS | 73.9 tok/s | 38.2 GB |
| Mixtral 8x7B InstructMistral AI | SS | 127.5 tok/s | 31.0 GB |
| Mistral Small 3 24BMistral AI | SS | 37.2 tok/s | 39.0 GB |
| Qwen3.6 35B-A3BAlibaba | SS | 169.8 tok/s | 27.1 GB |
| Qwen3.5-35B-A3BAlibaba | SS | 169.8 tok/s | 27.1 GB |
| MiniCPM-V 4.7 35B-A3Bopenbmb | SS | 169.8 tok/s | 27.2 GB |
| Apertus 70BEPFL, ETH Zurich, CSCS | SS | 33.4 tok/s | 43.4 GB |
KolibriAleph Alpha | 78B(3B active) | SS | 55.2 tok/s | 69.7 GB | |
| 81.29B(3B active) | SS | 169.8 tok/s | 53.9 GB | ||
| 74B(9B active) | SS | 240.9 tok/s | 43.7 GB | ||
| 74B(9B active) | SS | 240.9 tok/s | 43.7 GB | ||
Comet 60BMultiverse Computing | 61.1B(12.8B active) | SS | 173.7 tok/s | 36.4 GB | |
K2-Horizon-MoVA-36B-A4BInstitute of Foundation Models | 36B(4B active) | SS | 73.9 tok/s | 38.2 GB | |
Mixtral 8x7B InstructMistral AI | 46.7B(12.9B active) | SS | 127.5 tok/s | 31.0 GB | |
Mistral Small 3 24BMistral AI | 24B | SS | 37.2 tok/s | 39.0 GB | |
Ad Tools for AI BuildersAll the hardware and tools you need to run private AI at your office or at home.Shop Now | |||||
Qwen3.6 35B-A3BAlibaba | 35B(3B active) | SS | 169.8 tok/s | 27.1 GB | |
Qwen3.5-35B-A3BAlibaba | 35B(3B active) | SS | 169.8 tok/s | 27.1 GB | |
MiniCPM-V 4.7 35B-A3Bopenbmb | 35.21B(3B active) | SS | 169.8 tok/s | 27.2 GB | |
Apertus 70BEPFL, ETH Zurich, CSCS | 70B | SS | 33.4 tok/s | 43.4 GB | |
Llama 2 70B ChatMeta | 70B | SS | 33.4 tok/s | 43.4 GB | |
Xing4.0-29B-A4BChina Telecom AI | 29B(4B active) | SS | 129.3 tok/s | 25.7 GB | |
Gemma 4 12BGoogle | 12B | SS | 45.2 tok/s | 32.0 GB | |
Gemma 4 12B Coderyuxinlu1 | 12B | SS | 45.2 tok/s | 32.0 GB | |
Ad Vast.aiAffordable on-demand GPU rentals for training and inference. Pick from thousands of hosts.Rent a GPU | |||||
North Mini CodeCohere | 30B(3B active) | SS | 172.8 tok/s | 24.0 GB | |
Nemotron 3 Nano OmniNVIDIA | 30B(3B active) | SS | 169.8 tok/s | 24.2 GB | |
Gemma 4 26B-A4B ITGoogle | 26B(4B active) | SS | 131.6 tok/s | 23.8 GB | |
DiffusionGemma 26B-A4BGoogle | 25.2B(3.8B active) | SS | 138.2 tok/s | 22.9 GB | |
| 70B | SS | 31.7 tok/s | 45.7 GB | ||
LLaMA 65BMeta | 65B | SS | 36.9 tok/s | 39.3 GB | |
Qwen3-30B-A3BAlibaba | 30B(3B active) | SS | 269.0 tok/s | 21.0 GB | |
Holo4-35B-A3BHcompany | 35B(3B active) | SS | 619.7 tok/s | 20.9 GB | |
| AdTools for AI BuildersAll the hardware and tools you need to run private AI at your office or at home.Shop Now | |||||
GoTo-AI 27B InstructGoTo Group | 27B | SS | 33.1 tok/s | 43.8 GB | |
Model Compatibility
Use the calculator to weigh this device against any other, model by model, with speed and fit scores.
Straight answers to the questions we hear most often.
Can't find what you're looking for? Book a Discovery Call
Kolibri is the highest-rated model that runs on NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition at 4-bit, at about 55 tok/s. In total, NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition can run 280 of the 336 open models we track.
We size each model at 4-bit (Q4_K_M), add the KV cache for its context length and a fixed runtime overhead, then compare that to this device VRAM. Each model is graded on how comfortably it fits and how fast it should run on this card.
It is an estimate of decode speed in tokens per second at 4-bit, based on this device memory bandwidth and each model size. Real speed depends on your runtime, batch size, and context length, so use it to compare models, not as a hard promise.
Lower-bit quantization shrinks a model so more of them fit. These grades already assume 4-bit, which is the common local default. Going below 4-bit can squeeze a larger model in at some quality cost; going above needs more VRAM and may not fit.
Mixture-of-experts models only activate a fraction of their parameters at a time, so their runtime memory is closer to the active size than the headline size. We score fit on the active parameters, which is why some very large models still fit.
