readygpu

NVIDIA GeForce RTX 5090 vs NVIDIA RTX PRO 4500 Blackwell

For running models locally, priced in the UK. At 8k context the first runs 61 models fully in memory, the second 61.

In short

Both have 32 GB of memory for models; the NVIDIA GeForce RTX 5090 has 1,792 GB/s of bandwidth and the NVIDIA RTX PRO 4500 Blackwell 896 GB/s. At 8k context both run the same 61 models we track fully in memory. On the models both run fully in memory, the NVIDIA GeForce RTX 5090 is about 2.0× as fast as the NVIDIA RTX PRO 4500 Blackwell (estimated, median across models).

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

NVIDIA GeForce RTX 5090

£5,587

eBay

Add to build
NVIDIA RTX PRO 4500 Blackwell

Not priced here

Add to build

Specifications

SpecNVIDIA GeForce RTX 5090NVIDIA RTX PRO 4500 Blackwell
GPUGeForce RTX 5090RTX PRO 4500 Blackwell
Memory32 GB GDDR732 GB GDDR7
Memory bus512-bit256-bit
Memory bandwidth1,792 GB/s896 GB/s
FP16 compute209 TFLOPS54.94 TFLOPS
SoftwareCUDACUDA
Board power575 W200 W
PCIe generationGen 5Gen 5
PCIe lanesx16x16

What each runs

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

ModelNVIDIA GeForce RTX 5090NVIDIA RTX PRO 4500 Blackwell
Qwen 3.8 27B37tok/sQ8_019tok/sQ8_0
Qwen 3.6 27B38tok/sQ8_019tok/sQ8_0
Gemma 4 31B42tok/sQ6_K21tok/sQ6_K
Qwen 3.6 35B-A3B237tok/sQ6_K119tok/sQ6_K
Qwen 3.5 27B38tok/sQ8_019tok/sQ8_0
Gemma 4 26B-A4B147tok/sQ8_073tok/sQ8_0
Qwen 3.5 35B-A3B253tok/sQ6_K126tok/sQ6_K
GLM 4.7 Flash117tok/sQ8_058tok/sQ8_0
Devstral Small 2 24B42tok/sQ8_021tok/sQ8_0
Nemotron 3.5 Lightning 30B-A3B225tok/sUD-Q5_K_M112tok/sUD-Q5_K_M
Kimi Linear 48B-A3B343tok/sQ4_K_M172tok/sQ4_K_M
gpt-oss 20B253tok/sMXFP4127tok/sMXFP4
Qwen3 Coder 30B-A3B204tok/sQ6_K102tok/sQ6_K
Qwen3 32B38tok/sQ6_K19tok/sQ6_K
Gemma 4 12B85tok/sQ8_042tok/sQ8_0
Qwen 3.5 9B113tok/sQ8_056tok/sQ8_0
Mistral Small 3.2 24B42tok/sQ8_021tok/sQ8_0
DeepSeek R1 Distill Qwen 32B38tok/sQ6_K19tok/sQ6_K
Gemma 3 27B37tok/sQ8_019tok/sQ8_0
Ministral 3 14B71tok/sQ8_035tok/sQ8_0
Qwen3 14B65tok/sQ8_033tok/sQ8_0
DeepSeek R1 0528 Qwen3 8B112tok/sQ8_056tok/sQ8_0
Qwen3 8B112tok/sQ8_056tok/sQ8_0
Phi-4 14B64tok/sQ8_032tok/sQ8_0
Qwen 3.5 4B231tok/sQ8_0116tok/sQ8_0
Ministral 3 8B109tok/sQ8_055tok/sQ8_0
Granite 4.2 8B104tok/sQ8_052tok/sQ8_0
Gemma 4 E4B131tok/sQ8_066tok/sQ8_0
LFM2.5 8B-A1B455tok/sQ8_0227tok/sQ8_0
Llama 3.1 8B145tok/sQ6_K72tok/sQ6_K
Qwen2.5 0.5B1759tok/sQ8_0879tok/sQ8_0
Qwen3.5 0.8B1193tok/sQ8_0596tok/sQ8_0
Qwen3.5 2B715tok/sQ5_K_M357tok/sQ5_K_M
DeepSeek R1 Distill Qwen 1.5B522tok/sQ8_0261tok/sQ8_0
MiniCPM5 2B366tok/sQ8_0183tok/sQ8_0
LFM2.5 2.6B329tok/sQ8_0164tok/sQ8_0
LFM2.5 VL 3B329tok/sQ8_0164tok/sQ8_0
Spark X2.5 4B327tok/smlx-4bit163tok/smlx-4bit
granite vision 4.1 4b259tok/sQ8_0130tok/sQ8_0
Ministral 3 3B Instruct 2512246tok/sQ8_0123tok/sQ8_0
granite 4.2 3b243tok/sQ8_0122tok/sQ8_0
LFM2.5 24B A2B238tok/sQ8_0119tok/sQ8_0
Mellum2.1 12B A2.5B Thinking236tok/sQ8_0118tok/sQ8_0
FrogNano 4B 2609225tok/sQ8_0112tok/sQ8_0
Kimi VL A3B Thinking 2506199tok/sQ8_099tok/sQ8_0
Kimi VL A3B190tok/smlx-8bit95tok/smlx-8bit
DeepSeek R1 Distill Qwen 7B130tok/sQ8_065tok/sQ8_0
GLM 4 9B 0414129tok/smlx-5bit65tok/smlx-5bit
DeepSeek R1 Distill Llama 8B116tok/sQ8_058tok/sQ8_0
Ministral 3 8B Reasoning 2512109tok/sQ8_055tok/sQ8_0
GLM 4.6V Flash108tok/sQ8_054tok/sQ8_0
GLM 4.1V 9B Thinking92tok/smlx-8bit46tok/smlx-8bit
Ministral 3 14B Reasoning 251271tok/sQ8_035tok/sQ8_0
DeepSeek R1 Distill Qwen 14B64tok/sQ8_032tok/sQ8_0
Qwen3.8 27B OBLITERATED52tok/smlx-4bit26tok/smlx-4bit
Devstral Small 250742tok/sQ8_021tok/sQ8_0
Magistral Small 250642tok/sQ8_021tok/sQ8_0
Magistral Small 250742tok/sQ8_021tok/sQ8_0
Magistral Small 250942tok/sQ8_021tok/sQ8_0
granite 4.2 30b42tok/sQ6_K21tok/sQ6_K
Muse Glimmer 30B34tok/sQ8_017tok/sQ8_0
Qwen 3.5 122B-A10Boffloadoffload
Qwen3 Coder Nextoffloadoffload
Nemotron 3 Super 120B-A12Boffloadoffload
Mistral Small 4 119Boffloadoffload
GLM 4.5 Airoffloadoffload
Qwen3 Next 80B-A3Boffloadoffload
Llama 3.3 70Boffloadoffload
DeepSeek R1 Distill Llama 70Boffloadoffload
DeepSeek V4 Flashwon't runwon't run
GLM 5.3 Flashwon't runwon't run
Qwen 3.8 Flash Nextwon't runwon't run
MiniMax M2.7won't runwon't run
gpt-oss 120Bwon'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
Devstral 2 123B Instruct 2512won't runwon't run
GLM 4.5won't runwon't run
GLM 4.5Vwon't runwon't run
GLM 4.6won't runwon't run
GLM 4.6Vwon'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 Dev 72Bwon'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
Mistral Medium 3.5 128Bwon't runwon't run

Questions

Which is better for local AI, the NVIDIA GeForce RTX 5090 or the NVIDIA RTX PRO 4500 Blackwell?
Choose the NVIDIA GeForce RTX 5090 for speed.
Which can run bigger models, the NVIDIA GeForce RTX 5090 or the NVIDIA RTX PRO 4500 Blackwell?
At 8k context both run the same 61 models we track fully in memory.
Which is faster for local AI, the NVIDIA GeForce RTX 5090 or the NVIDIA RTX PRO 4500 Blackwell?
On the models both run fully in memory, the NVIDIA GeForce RTX 5090 is about 2.0× as fast as the NVIDIA RTX PRO 4500 Blackwell (estimated, median across models).

More comparisons

More hardware for local AI