readygpu

NVIDIA GeForce RTX 4090 vs NVIDIA RTX PRO 5000 Blackwell 72GB

For running models locally, priced in Canada. At 8k context the first runs 60 models fully in memory, the second 73.

In short

The NVIDIA RTX PRO 5000 Blackwell 72GB has more memory for models: 72 GB against 24 GB. At 8k context the NVIDIA RTX PRO 5000 Blackwell 72GB runs 73 of the models we track fully in memory and the NVIDIA GeForce RTX 4090 60; only the NVIDIA RTX PRO 5000 Blackwell 72GB holds Qwen 3.5 122B-A10B. On the models both run fully in memory, the NVIDIA RTX PRO 5000 Blackwell 72GB is about 1.3× as fast as the NVIDIA GeForce RTX 4090 (estimated, median across models). In Canada the NVIDIA GeForce RTX 4090 costs $3,300 at eBay and the NVIDIA RTX PRO 5000 Blackwell 72GB $13,470 at Amazon: the NVIDIA GeForce RTX 4090 is $10,170 cheaper.

Updated . Speeds are estimates; how they are worked out.

NVIDIA GeForce RTX 4090

$3,300

eBay

Add to build
NVIDIA RTX PRO 5000 Blackwell 72GB

$13,470

Amazon

Add to build

Specifications

SpecNVIDIA GeForce RTX 4090NVIDIA RTX PRO 5000 Blackwell 72GB
GPUGeForce RTX 4090RTX PRO 5000 Blackwell 72GB
Memory24 GB GDDR6X72 GB GDDR7
Memory bus384-bit384-bit
Memory bandwidth1,008 GB/s1,344 GB/s
FP16 compute165 TFLOPS73.69 TFLOPS
SoftwareCUDACUDA
Board power450 W300 W
PCIe generationGen 4Gen 5
PCIe lanesx16x16

What each runs

Estimated decode speed with the most precise quant that fits each one est.

ModelNVIDIA GeForce RTX 4090NVIDIA RTX PRO 5000 Blackwell 72GB
Qwen 3.8 27B28tok/sUD-Q6_K28tok/sQ8_0
Qwen 3.5 122B-A10Boffload103tok/sQ3_K_M
Qwen 3.6 27B27tok/sQ6_K29tok/sQ8_0
Gemma 4 31B27tok/sQ5_K_M24tok/sQ8_0
Qwen3 Coder Nextoffload144tok/sQ6_K
Nemotron 3 Super 120B-A12Bwon't run80tok/sUD-Q3_K_M
gpt-oss 120Bwon't run146tok/sMXFP4
Mistral Small 4 119Boffload110tok/sUD-Q4_K_M
Qwen 3.6 35B-A3B179tok/sQ4_K_M151tok/sQ8_0
Qwen 3.5 27B27tok/sQ6_K29tok/sQ8_0
Gemma 4 26B-A4B95tok/sUD-Q6_K110tok/sQ8_0
GLM 4.5 Airoffload47tok/sQ4_K_M
Qwen 3.5 35B-A3B181tok/sQ4_K_M151tok/sQ8_0
GLM 4.7 Flash94tok/sQ5_K_M87tok/sQ8_0
Devstral Small 2 24B30tok/sQ6_K32tok/sQ8_0
Qwen3 Next 80B-A3Boffload186tok/sQ6_K
Nemotron 3.5 Lightning 30B-A3B163tok/sMXFP4146tok/sUD-Q6_K_XL
gpt-oss 20B142tok/sMXFP4190tok/sMXFP4
Qwen3 Coder 30B-A3B128tok/sQ5_K_M125tok/sQ8_0
Kimi Linear 48B-A3Boffload152tok/sQ8_0
Qwen3 32B29tok/sQ4_K_M23tok/sQ8_0
Gemma 4 12B48tok/sQ8_063tok/sQ8_0
Qwen 3.5 9B64tok/sQ8_085tok/sQ8_0
Mistral Small 3.2 24B30tok/sQ6_K32tok/sQ8_0
DeepSeek R1 Distill Qwen 32B28tok/sQ4_K_M23tok/sQ8_0
Gemma 3 27B27tok/sQ6_K28tok/sQ8_0
Llama 3.3 70Boffload16tok/sQ5_K_M
Ministral 3 14B40tok/sQ8_053tok/sQ8_0
Qwen3 14B37tok/sQ8_049tok/sQ8_0
DeepSeek R1 0528 Qwen3 8B63tok/sQ8_084tok/sQ8_0
Qwen3 8B63tok/sQ8_084tok/sQ8_0
Phi-4 14B36tok/sQ8_048tok/sQ8_0
Qwen 3.5 4B130tok/sQ8_0174tok/sQ8_0
Ministral 3 8B62tok/sQ8_082tok/sQ8_0
Granite 4.2 8B59tok/sQ8_078tok/sQ8_0
Gemma 4 E4B74tok/sQ8_098tok/sQ8_0
LFM2.5 8B-A1B256tok/sQ8_0341tok/sQ8_0
Llama 3.1 8B82tok/sQ6_K109tok/sQ6_K
Qwen2.5 0.5B989tok/sQ8_01319tok/sQ8_0
Qwen3.5 0.8B671tok/sQ8_0895tok/sQ8_0
Qwen3.5 2B402tok/sQ5_K_M536tok/sQ5_K_M
DeepSeek R1 Distill Qwen 1.5B294tok/sQ8_0391tok/sQ8_0
MiniCPM5 2B206tok/sQ8_0275tok/sQ8_0
LFM2.5 2.6B185tok/sQ8_0247tok/sQ8_0
LFM2.5 VL 3B185tok/sQ8_0247tok/sQ8_0
Spark X2.5 4B184tok/smlx-4bit245tok/smlx-4bit
LFM2.5 24B A2B154tok/sQ6_K179tok/sQ8_0
granite vision 4.1 4b146tok/sQ8_0194tok/sQ8_0
Ministral 3 3B Instruct 2512138tok/sQ8_0184tok/sQ8_0
granite 4.2 3b137tok/sQ8_0183tok/sQ8_0
Mellum2.1 12B A2.5B Thinking133tok/sQ8_0177tok/sQ8_0
FrogNano 4B 2609127tok/sQ8_0169tok/sQ8_0
Kimi VL A3B Thinking 2506112tok/sQ8_0149tok/sQ8_0
Kimi VL A3B107tok/smlx-8bit142tok/smlx-8bit
DeepSeek R1 Distill Qwen 7B73tok/sQ8_097tok/sQ8_0
GLM 4 9B 041473tok/smlx-5bit97tok/smlx-5bit
DeepSeek R1 Distill Llama 8B65tok/sQ8_087tok/sQ8_0
Ministral 3 8B Reasoning 251262tok/sQ8_082tok/sQ8_0
GLM 4.6V Flash61tok/sQ8_081tok/sQ8_0
GLM 4.1V 9B Thinking52tok/smlx-8bit69tok/smlx-8bit
Ministral 3 14B Reasoning 251240tok/sQ8_053tok/sQ8_0
DeepSeek R1 Distill Qwen 14B36tok/sQ8_048tok/sQ8_0
Devstral Small 250735tok/sQ5_K_M32tok/sQ8_0
Magistral Small 250735tok/sQ5_K_M32tok/sQ8_0
Magistral Small 250935tok/sQ5_K_M32tok/sQ8_0
Magistral Small 250630tok/sQ6_K32tok/sQ8_0
Qwen3.8 27B OBLITERATED29tok/smlx-4bit39tok/smlx-4bit
granite 4.2 30b27tok/sQ5_K_M25tok/sQ8_0
Muse Glimmer 30B27tok/sQ6_K26tok/sQ8_0
DeepSeek R1 Distill Llama 70Boffload16tok/sQ5_K_M
GLM 4.5Vwon't run50tok/sQ4_K_M
GLM 4.6Vwon't run46tok/sQ4_K_M
Kimi Dev 72Bwon't run12tok/sQ6_K
DeepSeek V4 Flashwon't runoffload
Qwen 3.8 Flash Nextwon't runoffload
MiniMax M2.7won't runoffload
Devstral 2 123B Instruct 2512won't runoffload
Mistral Medium 3.5 128Bwon't runoffload
GLM 5.3 Flashwon't runwon't run
DeepSeek R1won't runwon't run
DeepSeek R1 0528won't runwon't run
DeepSeek R1 Zerowon't runwon't run
DeepSeek V3won't runwon't run
DeepSeek V3 0324won't runwon't run
DeepSeek V3.1won't runwon't run
DeepSeek V3.1 Terminuswon't runwon't run
DeepSeek V3.2won't runwon't run
DeepSeek V4 Flash 0731won't runwon't run
DeepSeek V4 Flash Vision Expwon't runwon't run
GLM 4.5won't runwon't run
GLM 4.6won't runwon't run
GLM 4.7won't runwon't run
GLM 5won't runwon't run
GLM 5.1won't runwon't run
GLM 5.2won't runwon't run
GLM 5.3won't runwon't run
Kimi K2won't runwon't run
Kimi K2 Instruct 0905won't runwon't run
Kimi K2 Thinkingwon't runwon't run
Kimi K2.5won't runwon't run
MiniMax M2won't runwon't run
MiniMax M2.1won't runwon't run
MiniMax M2.5won't runwon't run
MiniMax M3won't runwon't run

Questions

Which is better for local AI, the NVIDIA GeForce RTX 4090 or the NVIDIA RTX PRO 5000 Blackwell 72GB?
Choose the NVIDIA RTX PRO 5000 Blackwell 72GB to run bigger models. Choose the NVIDIA RTX PRO 5000 Blackwell 72GB for speed. Choose the NVIDIA GeForce RTX 4090 to spend less.
Which can run bigger models, the NVIDIA GeForce RTX 4090 or the NVIDIA RTX PRO 5000 Blackwell 72GB?
At 8k context the NVIDIA RTX PRO 5000 Blackwell 72GB runs 73 of the models we track fully in memory and the NVIDIA GeForce RTX 4090 60; only the NVIDIA RTX PRO 5000 Blackwell 72GB holds Qwen 3.5 122B-A10B.
Which is faster for local AI, the NVIDIA GeForce RTX 4090 or the NVIDIA RTX PRO 5000 Blackwell 72GB?
On the models both run fully in memory, the NVIDIA RTX PRO 5000 Blackwell 72GB is about 1.3× as fast as the NVIDIA GeForce RTX 4090 (estimated, median across models).
What is the price difference between the NVIDIA GeForce RTX 4090 and the NVIDIA RTX PRO 5000 Blackwell 72GB?
In Canada the NVIDIA GeForce RTX 4090 costs $3,300 at eBay and the NVIDIA RTX PRO 5000 Blackwell 72GB $13,470 at Amazon: the NVIDIA GeForce RTX 4090 is $10,170 cheaper.

More comparisons

More hardware for local AI