NVIDIA RTX PRO 4000 Blackwell
24 GB · 672 GB/s · 145 W
In short
The NVIDIA RTX PRO 4000 Blackwell has 24 GB of GDDR7 memory with 672 GB/s of bandwidth, which sets how fast it generates text. At 128k context it runs 31 of the 125 open models we track fully in memory; the most capable is Gemma 4 26B-A4B at Q5_K_M, at about 43 tok/s (estimated). We have no current price for the NVIDIA RTX PRO 4000 Blackwell in the US.
Updated . Speeds are estimates; how they are worked out.
Prices
Every retailer we track in the US, cheapest first. Buy links may earn us a commission.
No retailer in this market lists it right now.
What it runs
Every model in the catalogue, with the best quant for this hardware. Estimates, not measurements.
| Model | Quant | Fits | Memory | Max context | Decode @128k | Prefill | Good for |
|---|---|---|---|---|---|---|---|
| Gemma 4 26B-A4B | Q5_K_M | fits | 22.1 GB | 177k256k with q8 cache | 43tok/sest. | 1531 tok/s | chat, coding, agents, long documents |
| gpt-oss 20B OpenAI | MXFP4 | fits | 15.7 GB | 128k | 44tok/sest. | 991 tok/s | chat, coding, agents, long documents |
| Gemma 4 12B | Q8_0 | fits | 15.6 GB | 256k | 28tok/sest. | 636 tok/s | chat, coding, long documents |
| Qwen 3.5 9B Alibaba Qwen | Q8_0 | fits | 14.4 GB | 256k | 30tok/sest. | 757 tok/s | chat, coding, long documents |
| Qwen 3.5 4B Alibaba Qwen | Q8_0 | fits | 9.7 GB | 256k | 47tok/sest. | 1178 tok/s | chat, coding, long documents |
| Ministral 3 8B Mistral AI | Q4_K_M | fits | 23.3 GB | 128k233k with q8 cache | 18tok/sest. | 388 tok/s | chat, coding, long documents |
| Gemma 4 E4B | Q8_0 | fits | 12.6 GB | 128k | 35tok/sest. | 1061 tok/s | chat, coding, long documents |
| LFM2.5 8B-A1B Liquid AI | Q8_0 | fits | 15.8 GB | 125k | 33tok/sest. | 1418 tok/s | chat, coding, long documents |
| Llama 3.1 8B Meta | Q5_K_M | fits | 22.8 GB | 128k | 18tok/sest. | 419 tok/s | chat, coding, long documents |
| Qwen2.5 0.5B Alibaba Qwen | Q8_0 | fits | 3.4 GB | 32k | 195tok/sest. | 3186 tok/s | chat, coding, long documents |
| Qwen3.5 0.8B Alibaba Qwen | Q8_0 | fits | 3.7 GB | 256k | 171tok/sest. | 4248 tok/s | chat, coding, agents, long documents |
| Qwen3.5 2B Alibaba Qwen | Q5_K_M | fits | 4.3 GB | 256k | 136tok/sest. | 2716 tok/s | chat, coding, agents, long documents |
| DeepSeek R1 Distill Qwen 1.5B DeepSeek | Q8_0 | fits | 6.7 GB | 128k | 74tok/sest. | 1423 tok/s | chat, coding, long documents |
| Mellum2.1 12B A2.5B Thinking JetBrains | Q8_0 | fits | 15.3 GB | 128k | 55tok/sest. | 1640 tok/s | chat, coding, agents, long documents |
| Spark X2.5 4B XHToken | mlx-4bit | fits | 8.8 GB | 483k808k with q8 cache | 53tok/sest. | 1165 tok/s | chat, coding, agents, long documents |
| MiniCPM5 2B openbmb | Q8_0 | fits | 9.2 GB | 128k | 50tok/sest. | 765 tok/s | chat, coding, agents, long documents |
| FrogNano 4B 2609 Microsoft | Q8_0 | fits | 9.8 GB | 256k | 47tok/sest. | 1178 tok/s | chat, coding, agents, long documents |
| LFM2.5 2.6B Liquid AI | Q8_0 | fits | 11.6 GB | 128k | 38tok/sest. | 982 tok/s | chat, coding, agents, long documents |
| LFM2.5 VL 3B Liquid AI | Q8_0 | fits | 11.6 GB | 32k | 38tok/sest. | 944 tok/s | chat, coding, agents, long documents |
| Kimi VL A3B Thinking 2506 Moonshot AI | Q8_0 | fits | 21.1 GB | 128k | 35tok/sest. | 1030 tok/s | chat, coding, long documents |
| Kimi VL A3B Moonshot AI | mlx-8bit | fits | 21.8 GB | 128k | 34tok/sest. | 1030 tok/s | chat, coding, agents, long documents |
| GLM 4 9B 0414 Z.ai | mlx-5bit | fits | 14.1 GB | 32k | 31tok/sest. | 342 tok/s | chat, coding, agents, long documents |
| granite vision 4.1 4b IBM | Q8_0 | fits | 14.8 GB | 128k | 29tok/sest. | 606 tok/s | chat, coding, long documents |
| granite 4.2 3b IBM | Q8_0 | fits | 15.1 GB | 128k | 29tok/sest. | 618 tok/s | chat, coding, agents, long documents |
| GLM 4.6V Flash Z.ai | Q8_0 | fits | 15.8 GB | 128k | 27tok/sest. | 332 tok/s | chat, coding, agents, long documents |
| DeepSeek R1 Distill Qwen 7B DeepSeek | Q8_0 | fits | 16.0 GB | 128k | 27tok/sest. | 508 tok/s | chat, coding, long documents |
| GLM 4.1V 9B Thinking Z.ai | mlx-8bit | fits | 17.4 GB | 64k | 24tok/sest. | 332 tok/s | chat, coding, long documents |
| Ministral 3 3B Instruct 2512 Mistral AI | Q8_0 | fits | 17.9 GB | 179k256k with q8 cache | 24tok/sest. | 593 tok/s | chat, coding, agents, long documents |
| Muse Glimmer 30B meta-models | Q5_K_M | fits | 21.9 GB | 128k | 19tok/sest. | 282 tok/s | chat, coding |
| DeepSeek R1 Distill Llama 8B DeepSeek | Q5_K_M | fits | 22.8 GB | 128k | 18tok/sest. | 419 tok/s | chat, coding, long documents |
| Ministral 3 8B Reasoning 2512 Mistral AI | Q4_K_M | fits | 23.3 GB | 128k233k with q8 cache | 18tok/sest. | 388 tok/s | chat, coding, agents, long documents |
| Qwen 3.8 27B Alibaba Qwen | UD-Q4_K_M | offload | 24.9 GB | 105k184k with q8 cache | 12tok/sest. | 249 tok/s | chat, coding |
| Qwen 3.5 122B-A10B Alibaba Qwen | UD-Q2_K_XL | offload | 43.6 GB | – | 9tok/sest. | 329 tok/s | chat |
| Qwen 3.6 27B Alibaba Qwen | Q4_K_M | offload | 25.3 GB | 100k176k with q8 cache | 11tok/sest. | 247 tok/s | chat, coding |
| Gemma 4 31B | Q4_K_M | offload | 29.3 GB | 57k110k with q8 cache | 6tok/sest. | 215 tok/s | chat |
| Mistral Small 4 119B Mistral AI | UD-Q2_K_XL | offload | 41.7 GB | – | 11tok/sest. | 291 tok/s | chat, coding |
| Qwen 3.6 35B-A3B Alibaba Qwen | MXFP4 | offload | 24.2 GB | 95k149k with q8 cache | 44tok/sest. | 1234 tok/s | chat, coding, agents, long documents |
| Qwen 3.5 27B Alibaba Qwen | Q4_K_M | offload | 25.2 GB | 101k178k with q8 cache | 11tok/sest. | 248 tok/s | chat, coding |
| Qwen 3.5 35B-A3B Alibaba Qwen | Q4_K_M | offload | 24.5 GB | 84k131k with q8 cache | 41tok/sest. | 1224 tok/s | chat, coding, agents, long documents |
| GLM 4.7 Flash Z.ai | Q4_K_M | offload | 25.1 GB | 98k169k with q8 cache | 17tok/sest. | 569 tok/s | chat, coding, agents, long documents |
| Devstral Small 2 24B Mistral AI | Q4_K_M | offload | 34.8 GB | 57k105k with q8 cache | 4tok/sest. | 184 tok/s | chat, coding |
| Nemotron 3.5 Lightning 30B-A3B NVIDIA | MXFP4 | offload | 23.9 GB | 88k115k with q8 cache | 72tok/sest. | 1675 tok/s | chat, coding, agents, long documents |
| Kimi Linear 48B-A3B Moonshot AI | Q4_K_M | offload | 30.5 GB | – | 32tok/sest. | 808 tok/s | chat, coding, long documents |
| Qwen3 Coder 30B-A3B Alibaba Qwen | Q4_K_M | offload | 30.7 GB | 54k97k with q8 cache | 6tok/sest. | 312 tok/s | chat, coding, agents |
| Gemma 3 27B | Q4_K_M | offload | 27.3 GB | 81k128k with q8 cache | 8tok/sest. | 289 tok/s | chat, coding |
| Mistral Small 3.2 24B Mistral AI | Q4_K_M | offload | 34.8 GB | 57k105k with q8 cache | 4tok/sest. | 184 tok/s | chat, coding |
| Ministral 3 14B Mistral AI | Q4_K_M | offload | 29.1 GB | 92k169k with q8 cache | 6tok/sest. | 261 tok/s | chat, coding |
| Qwen3 14B Alibaba Qwen | Q4_K_M | offload | 29.8 GB | 40k | 5tok/sest. | 219 tok/s | chat, coding |
| DeepSeek R1 0528 Qwen3 8B DeepSeek | Q4_K_M | offload | 24.1 GB | 122k128k with q8 cache | 14tok/sest. | 376 tok/s | chat, coding |
| Qwen3 8B Alibaba Qwen | Q4_K_M | offload | 24.1 GB | 40k | 14tok/sest. | 376 tok/s | chat, coding |
| Phi-4 14B Microsoft | Q4_K_M | offload | 34.7 GB | 16k | 4tok/sest. | 202 tok/s | chat, coding |
| Granite 4.2 8B IBM | Q4_K_M | offload | 26.6 GB | 108k128k with q8 cache | 8tok/sest. | 322 tok/s | chat, coding, agents |
| LFM2.5 24B A2B Liquid AI | Q4_K_M | offload | 25.7 GB | 100k128k with q8 cache | 13tok/sest. | 780 tok/s | chat, coding, agents |
| Qwen3.8 27B OBLITERATED OBLITERATUS | mlx-4bit | offload | 28.8 GB | 48k84k with q8 cache | 6tok/sest. | 229 tok/s | chat |
| Ministral 3 14B Reasoning 2512 Mistral AI | Q4_K_M | offload | 29.1 GB | 92k169k with q8 cache | 6tok/sest. | 261 tok/s | chat, coding, agents |
| DeepSeek R1 Distill Qwen 14B DeepSeek | Q4_K_M | offload | 33.8 GB | 74k128k with q8 cache | 4tok/sest. | 181 tok/s | chat, coding |
| Devstral Small 2507 Mistral AI | Q4_K_M | offload | 34.8 GB | 57k105k with q8 cache | 4tok/sest. | 186 tok/s | chat, coding |
| Magistral Small 2506 Mistral AI | Q4_K_M | offload | 34.8 GB | 40k | 4tok/sest. | 186 tok/s | chat, coding |
| Magistral Small 2507 Mistral AI | Q4_K_M | offload | 34.8 GB | 40k | 4tok/sest. | 186 tok/s | chat, coding |
| Magistral Small 2509 Mistral AI | Q4_K_M | offload | 34.8 GB | 57k105k with q8 cache | 4tok/sest. | 184 tok/s | chat, coding |
| DeepSeek V4 Flash DeepSeek | UD-Q2_K_XL | won't run | 102.4 GB | – | – | – | – |
| GLM 5.3 Flash Z.ai | UD-Q2_K_XL | won't run | 104.2 GB | – | – | – | – |
| Qwen 3.8 Flash Next Alibaba Qwen | UD-Q2_K_XL | won't run | 78.0 GB | – | – | – | – |
| MiniMax M2.7 MiniMax | UD-Q2_K_XL | won't run | 102.6 GB | – | – | – | – |
| Nemotron 3 Super 120B-A12B NVIDIA | UD-Q2_K_XL | won't run | 53.5 GB | – | – | – | – |
| Qwen3 Coder Next Alibaba Qwen | Q4_K_M | won't run | 49.7 GB | – | – | – | – |
| gpt-oss 120B OpenAI | Q4_K_M | won't run | 64.4 GB | – | – | – | – |
| GLM 4.5 Air Z.ai | Q2_K | won't run | 66.7 GB | – | – | – | – |
| Qwen3 Next 80B-A3B Alibaba Qwen | Q4_K_M | won't run | 49.7 GB | – | – | – | – |
| Qwen3 32B Alibaba Qwen | Q4_K_M | won't run | 51.9 GB | 16k30k with q8 cache | – | – | – |
| DeepSeek R1 Distill Qwen 32B DeepSeek | Q4_K_M | won't run | 51.9 GB | 16k30k with q8 cache | – | – | – |
| Llama 3.3 70B Meta | Q4_K_M | won't run | 81.1 GB | – | – | – | – |
| DeepSeek R1 DeepSeek | Q4_K_M | won't run | 386.7 GB | – | – | – | – |
| DeepSeek R1 0528 DeepSeek | Q4_K_M | won't run | 387.2 GB | – | – | – | – |
| DeepSeek R1 Distill Llama 70B DeepSeek | Q4_K_M | won't run | 81.1 GB | – | – | – | – |
| DeepSeek R1 Zero DeepSeek | Q4_K_M | won't run | 386.7 GB | – | – | – | – |
| DeepSeek V3 DeepSeek | Q4_K_M | won't run | 386.7 GB | – | – | – | – |
| DeepSeek V3 0324 DeepSeek | Q4_K_M | won't run | 386.7 GB | – | – | – | – |
| DeepSeek V3.1 DeepSeek | Q4_K_M | won't run | 387.6 GB | – | – | – | – |
| DeepSeek V3.1 Terminus DeepSeek | Q4_K_M | won't run | 387.6 GB | – | – | – | – |
| DeepSeek V3.2 DeepSeek | Q4_K_M | won't run | 387.6 GB | – | – | – | – |
| DeepSeek V4 Flash 0731 DeepSeek | MXFP4 | won't run | 157.8 GB | – | – | – | – |
| DeepSeek V4 Flash Vision Exp DeepSeek | MXFP4 | won't run | 166.6 GB | – | – | – | – |
| Devstral 2 123B Instruct 2512 Mistral AI | Q4_K_M | won't run | 115.2 GB | – | – | – | – |
| GLM 4.5 Z.ai | Q4_K_M | won't run | 249.0 GB | – | – | – | – |
| GLM 4.5V Z.ai | Q4_K_M | won't run | 83.7 GB | – | – | – | – |
| GLM 4.6 Z.ai | Q4_K_M | won't run | 248.2 GB | – | – | – | – |
| GLM 4.6V Z.ai | Q4_K_M | won't run | 90.1 GB | – | – | – | – |
| GLM 4.7 Z.ai | Q4_K_M | won't run | 249.0 GB | – | – | – | – |
| GLM 5 Z.ai | MXFP4 | won't run | 394.8 GB | – | – | – | – |
| GLM 5.1 Z.ai | MXFP4 | won't run | 432.4 GB | – | – | – | – |
| GLM 5.2 Z.ai | UD-Q4_K_M | won't run | 446.3 GB | – | – | – | – |
| GLM 5.3 Z.ai | Q8_0 | won't run | 758.7 GB | – | – | – | – |
| Kimi Dev 72B Moonshot AI | Q6_K | won't run | 101.4 GB | – | – | – | – |
| Kimi K2 Moonshot AI | Q4_K_M | won't run | 588.2 GB | – | – | – | – |
| Kimi K2 Instruct 0905 Moonshot AI | Q4_K_M | won't run | 588.6 GB | – | – | – | – |
| Kimi K2 Thinking Moonshot AI | Q4_K_M | won't run | 588.6 GB | – | – | – | – |
| Kimi K2.5 Moonshot AI | Q4_K_M | won't run | 588.6 GB | – | – | – | – |
| MiniMax M2 MiniMax | Q4_K_M | won't run | 161.3 GB | – | – | – | – |
| MiniMax M2.1 MiniMax | Q4_K_M | won't run | 161.3 GB | – | – | – | – |
| MiniMax M2.5 MiniMax | MXFP4 | won't run | 147.7 GB | – | – | – | – |
| MiniMax M3 MiniMax | MXFP4 | won't run | 256.1 GB | – | – | – | – |
| Mistral Medium 3.5 128B Mistral AI | Q4_K_M | won't run | 115.2 GB | – | – | – | – |
| granite 4.2 30b IBM | Q4_K_M | won't run | 50.2 GB | 23k42k with q8 cache | – | – | – |
Specifications
From the manufacturer's published specifications; sources are linked where we have them.
- GPU
- RTX PRO 4000 Blackwell
- Memory
- 24 GB GDDR7
- Memory bus
- 192-bit
- Memory bandwidth
- 672 GB/s
- FP16 compute
- 46.9 TFLOPS
- Software
- CUDA
- Board power
- 145 W
- PCIe generation
- Gen 5
- PCIe lanes
- x16
Compare with
Questions
- How much memory does the NVIDIA RTX PRO 4000 Blackwell have for AI models?
- The NVIDIA RTX PRO 4000 Blackwell has 24 GB of GDDR7 memory with 672 GB/s of bandwidth, which sets how fast it generates text.
- What is the best model the NVIDIA RTX PRO 4000 Blackwell can run locally?
- Gemma 4 26B-A4B (Q5_K_M) fits fully in its memory at 128k context and generates at about 43 tok/s, an estimate from its memory bandwidth. 31 of the 125 models we track fit at that context.
- Can the NVIDIA RTX PRO 4000 Blackwell run DeepSeek V4 Flash?
- No: even at UD-Q2_K_XL, DeepSeek V4 Flash needs about 102 GB at 128k context, more than the NVIDIA RTX PRO 4000 Blackwell and system RAM can hold.
- How much does the NVIDIA RTX PRO 4000 Blackwell cost in the US?
- We have no current price for the NVIDIA RTX PRO 4000 Blackwell in the US.
Speeds on this page are for 128k context. See the best hardware for local AI at every budget.