readygpu

NVIDIA GeForce RTX 4090 vs NVIDIA GeForce RTX 5090

For running models locally, priced in the UK. At 64k context the first runs 40 models fully in memory, the second 49.

In short

The NVIDIA GeForce RTX 5090 has more memory for models: 32 GB against 24 GB. At 64k context the NVIDIA GeForce RTX 5090 runs 49 of the models we track fully in memory and the NVIDIA GeForce RTX 4090 40; only the NVIDIA GeForce RTX 5090 holds Gemma 4 31B. On the models both run fully in memory, the NVIDIA GeForce RTX 5090 is about 1.8× as fast as the NVIDIA GeForce RTX 4090 (estimated, median across models). In the UK the NVIDIA GeForce RTX 4090 costs £6,015 at Amazon and the NVIDIA GeForce RTX 5090 £5,000 at eBay: the NVIDIA GeForce RTX 5090 is £1,015 cheaper.

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

NVIDIA GeForce RTX 4090

£6,015

Amazon

Add to build
NVIDIA GeForce RTX 5090

£5,000

eBay

Add to build

Specifications

SpecNVIDIA GeForce RTX 4090NVIDIA GeForce RTX 5090
GPUGeForce RTX 4090GeForce RTX 5090
Memory24 GB GDDR6X32 GB GDDR7
Memory bus384-bit512-bit
Memory bandwidth1,008 GB/s1,792 GB/s
FP16 compute165 TFLOPS209 TFLOPS
SoftwareCUDACUDA
Board power450 W575 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 GeForce RTX 5090
Qwen 3.8 27B29tok/sQ4_K_M39tok/sQ6_K
Qwen 3.6 27B26tok/sQ5_K_M41tok/sQ6_K
Gemma 4 31Boffload35tok/sQ6_K
Qwen 3.6 35B-A3B115tok/sQ4_K_M167tok/sQ6_K
Qwen 3.5 27B30tok/sQ4_K_M41tok/sQ6_K
Gemma 4 26B-A4B78tok/sUD-Q5_K_M117tok/sQ8_0
Qwen 3.5 35B-A3B116tok/sQ4_K_M175tok/sQ6_K
GLM 4.7 Flash57tok/sQ4_K_M87tok/sQ6_K
Devstral Small 2 24Boffload37tok/sQ6_K
gpt-oss 20B93tok/sMXFP4165tok/sMXFP4
Kimi Linear 48B-A3Boffload277tok/sQ4_K_M
Qwen3 Coder 30B-A3Boffload75tok/sQ6_K
Gemma 4 12B44tok/sQ8_079tok/sQ8_0
Qwen 3.5 9B53tok/sQ8_095tok/sQ8_0
Gemma 3 27B28tok/sQ4_K_M40tok/sQ6_K
Mistral Small 3.2 24Boffload37tok/sQ6_K
Ministral 3 14B29tok/sQ6_K44tok/sQ8_0
Qwen3 14B27tok/sQ6_K42tok/sQ8_0
DeepSeek R1 0528 Qwen3 8B34tok/sQ8_061tok/sQ8_0
Qwen3 8B34tok/sQ8_061tok/sQ8_0
Phi-4 14B26tok/sQ5_K_M38tok/sQ8_0
Qwen 3.5 4B94tok/sQ8_0166tok/sQ8_0
Ministral 3 8B34tok/sQ8_061tok/sQ8_0
Granite 4.2 8B31tok/sQ8_055tok/sQ8_0
Gemma 4 E4B62tok/sQ8_0110tok/sQ8_0
LFM2.5 8B-A1B88tok/sQ8_0156tok/sQ8_0
Llama 3.1 8B41tok/sQ6_K73tok/sQ6_K
Qwen2.5 0.5B468tok/sQ8_0831tok/sQ8_0
Qwen3.5 0.8B382tok/sQ8_0679tok/sQ8_0
Qwen3.5 2B277tok/sQ5_K_M492tok/sQ5_K_M
DeepSeek R1 Distill Qwen 1.5B166tok/sQ8_0294tok/sQ8_0
MiniCPM5 2B114tok/sQ8_0202tok/sQ8_0
Mellum2.1 12B A2.5B Thinking103tok/sQ8_0183tok/sQ8_0
LFM2.5 2.6B91tok/sQ8_0161tok/sQ8_0
LFM2.5 VL 3B91tok/sQ8_0161tok/sQ8_0
Kimi VL A3B71tok/smlx-8bit126tok/smlx-8bit
granite vision 4.1 4b70tok/sQ8_0124tok/sQ8_0
granite 4.2 3b68tok/sQ8_0120tok/sQ8_0
LFM2.5 24B A2B54tok/sQ5_K_M89tok/sQ8_0
DeepSeek R1 Distill Qwen 7B53tok/sQ8_094tok/sQ8_0
GLM 4.6V Flash49tok/sQ8_088tok/sQ8_0
GLM 4.1V 9B Thinking43tok/smlx-8bit77tok/smlx-8bit
DeepSeek R1 Distill Llama 8B37tok/sQ8_065tok/sQ8_0
Ministral 3 8B Reasoning 251234tok/sQ8_061tok/sQ8_0
DeepSeek R1 Distill Qwen 14B27tok/sQ5_K_M39tok/sQ8_0
Devstral Small 2507offload40tok/sQ5_K_M
Magistral Small 2506offload37tok/sQ6_K
Magistral Small 2507offload40tok/sQ5_K_M
Magistral Small 2509offload40tok/sQ5_K_M
Qwen 3.5 122B-A10Boffloadoffload
Qwen3 Coder Nextoffloadoffload
Nemotron 3 Super 120B-A12Bwon't runoffload
Mistral Small 4 119Boffloadoffload
GLM 4.5 Airwon't runoffload
Qwen3 Next 80B-A3Boffloadoffload
Nemotron 3.5 Lightning 30B-A3Boffloadoffload
Qwen3 32Boffloadoffload
DeepSeek R1 Distill Qwen 32Boffloadoffload
granite 4.2 30boffloadoffload
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
Llama 3.3 70Bwon't runwon't run
DeepSeek R1won't runwon't run
DeepSeek R1 0528won't runwon't run
DeepSeek R1 Distill Llama 70Bwon't runwon't run
DeepSeek V3 0324won't runwon't run
DeepSeek V3.1won'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 4.5Vwon't runwon't run
GLM 4.6Vwon't runwon't run
GLM 4.7won'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
Mistral Medium 3.5 128Bwon't runwon't run

Questions

Which is better for local AI, the NVIDIA GeForce RTX 4090 or the NVIDIA GeForce RTX 5090?
Choose the NVIDIA GeForce RTX 5090 to run bigger models. Choose the NVIDIA GeForce RTX 5090 for speed. Choose the NVIDIA GeForce RTX 5090 to spend less.
Which can run bigger models, the NVIDIA GeForce RTX 4090 or the NVIDIA GeForce RTX 5090?
At 64k context the NVIDIA GeForce RTX 5090 runs 49 of the models we track fully in memory and the NVIDIA GeForce RTX 4090 40; only the NVIDIA GeForce RTX 5090 holds Gemma 4 31B.
Which is faster for local AI, the NVIDIA GeForce RTX 4090 or the NVIDIA GeForce RTX 5090?
On the models both run fully in memory, the NVIDIA GeForce RTX 5090 is about 1.8× 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 GeForce RTX 5090?
In the UK the NVIDIA GeForce RTX 4090 costs £6,015 at Amazon and the NVIDIA GeForce RTX 5090 £5,000 at eBay: the NVIDIA GeForce RTX 5090 is £1,015 cheaper.

More comparisons