readygpu

NVIDIA GeForce RTX 3090 vs NVIDIA GeForce RTX 4060

For running models locally, priced in the UK. At 64k context the first runs 49 models fully in memory, the second 11.

In short

The NVIDIA GeForce RTX 3090 has more memory for models: 24 GB against 8 GB. At 64k context the NVIDIA GeForce RTX 3090 runs 49 of the models we track fully in memory and the NVIDIA GeForce RTX 4060 11; only the NVIDIA GeForce RTX 3090 holds Qwen 3.8 27B. On the models both run fully in memory, the NVIDIA GeForce RTX 3090 is about 3.4× as fast as the NVIDIA GeForce RTX 4060 (estimated, median across models). In the UK the NVIDIA GeForce RTX 3090 costs £1,330 at Amazon and the NVIDIA GeForce RTX 4060 £290 at eBay (used): the NVIDIA GeForce RTX 4060 is £1,040 cheaper.

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

NVIDIA GeForce RTX 3090

£1,330

Amazon

Add to build
NVIDIA GeForce RTX 4060

£290

eBay, used (asking)

Add to build

Specifications

SpecNVIDIA GeForce RTX 3090NVIDIA GeForce RTX 4060
GPUGeForce RTX 3090GeForce RTX 4060
Memory24 GB GDDR6X8 GB GDDR6
Memory bus384-bit128-bit
Memory bandwidth936 GB/s272 GB/s
FP16 compute71 TFLOPS15.114 TFLOPS
SoftwareCUDACUDA
Board power350 W115 W
PCIe generationGen 4Gen 4
PCIe lanesx16x8

What each runs

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

ModelNVIDIA GeForce RTX 3090NVIDIA GeForce RTX 4060
Qwen 3.8 27B27tok/sQ4_K_Moffload
Qwen 3.6 27B24tok/sQ5_K_Moffload
Qwen 3.6 35B-A3B107tok/sQ4_K_Moffload
Qwen 3.5 27B27tok/sQ4_K_Moffload
Gemma 4 26B-A4B72tok/sUD-Q5_K_Moffload
Qwen 3.5 35B-A3B107tok/sQ4_K_Moffload
GLM 4.7 Flash53tok/sQ4_K_Moffload
Nemotron 3.5 Lightning 30B-A3B131tok/sMXFP4offload
gpt-oss 20B86tok/sMXFP4offload
Gemma 4 12B41tok/sQ8_0offload
Qwen 3.5 9B50tok/sQ8_0offload
Gemma 3 27B26tok/sQ4_K_Moffload
Ministral 3 14B27tok/sQ6_Koffload
Qwen3 14B25tok/sQ6_Koffload
DeepSeek R1 0528 Qwen3 8B32tok/sQ8_0offload
Qwen3 8B32tok/sQ8_0offload
Phi-4 14B24tok/sQ5_K_Moffload
Qwen 3.5 4B87tok/sQ8_025tok/sQ8_0
Ministral 3 8B32tok/sQ8_0offload
Granite 4.2 8B29tok/sQ8_0offload
Gemma 4 E4B57tok/sQ8_0offload
LFM2.5 8B-A1B81tok/sQ8_0offload
Llama 3.1 8B38tok/sQ6_Koffload
Qwen2.5 0.5B434tok/sQ8_0126tok/sQ8_0
Qwen3.5 0.8B355tok/sQ8_0103tok/sQ8_0
Qwen3.5 2B257tok/sQ5_K_M75tok/sQ5_K_M
LFM3 1.7B218tok/sQ8_063tok/sQ8_0
DeepSeek R1 Distill Qwen 1.5B154tok/sQ8_045tok/sQ8_0
MiniCPM5 2B106tok/sQ8_031tok/sQ8_0
Spark X2.5 4B105tok/smlx-4bit31tok/smlx-4bit
Mellum2.1 12B A2.5B Thinking96tok/sQ8_0offload
FrogNano 4B 260985tok/sQ8_028tok/sQ6_K
LFM2.5 2.6B84tok/sQ8_027tok/sQ6_K
LFM2.5 VL 3B84tok/sQ8_027tok/sQ6_K
Kimi VL A3B Thinking 250668tok/sQ8_0offload
Kimi VL A3B66tok/smlx-8bitoffload
granite vision 4.1 4b65tok/sQ8_0offload
granite 4.2 3b63tok/sQ8_0offload
Ministral 3 3B Instruct 251255tok/sQ8_0offload
GLM 4 9B 041453tok/smlx-5bitoffload
LFM2.5 24B A2B50tok/sQ5_K_Moffload
DeepSeek R1 Distill Qwen 7B49tok/sQ8_0offload
GLM 4.6V Flash46tok/sQ8_0offload
GLM 4.1V 9B Thinking40tok/smlx-8bitoffload
DeepSeek R1 Distill Llama 8B34tok/sQ8_0offload
Ministral 3 8B Reasoning 251232tok/sQ8_0offload
Muse Glimmer 30B28tok/sQ5_K_Moffload
Ministral 3 14B Reasoning 251227tok/sQ6_Koffload
DeepSeek R1 Distill Qwen 14B25tok/sQ5_K_Moffload
Qwen 3.5 122B-A10Boffloadwon't run
Gemma 4 31Boffloadoffload
Qwen3 Coder Nextoffloadwon't run
Mistral Small 4 119Boffloadwon't run
Qwen3 Next 80B-A3Boffloadwon't run
Devstral Small 2 24Boffloadoffload
Kimi Linear 48B-A3Boffloadoffload
Qwen3 Coder 30B-A3Boffloadoffload
Qwen3 32Boffloadwon't run
Mistral Small 3.2 24Boffloadoffload
DeepSeek R1 Distill Qwen 32Boffloadwon't run
Devstral Small 2507offloadoffload
Magistral Small 2506offloadoffload
Magistral Small 2507offloadoffload
Magistral Small 2509offloadoffload
Qwen3.8 27B OBLITERATEDoffloadoffload
granite 4.2 30boffloadwon't run
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
Nemotron 3 Super 120B-A12Bwon't runwon't run
gpt-oss 120Bwon't runwon't run
GLM 4.5 Airwon'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 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 3090 or the NVIDIA GeForce RTX 4060?
Choose the NVIDIA GeForce RTX 3090 to run bigger models. Choose the NVIDIA GeForce RTX 3090 for speed. Choose the NVIDIA GeForce RTX 4060 to spend less.
Which can run bigger models, the NVIDIA GeForce RTX 3090 or the NVIDIA GeForce RTX 4060?
At 64k context the NVIDIA GeForce RTX 3090 runs 49 of the models we track fully in memory and the NVIDIA GeForce RTX 4060 11; only the NVIDIA GeForce RTX 3090 holds Qwen 3.8 27B.
Which is faster for local AI, the NVIDIA GeForce RTX 3090 or the NVIDIA GeForce RTX 4060?
On the models both run fully in memory, the NVIDIA GeForce RTX 3090 is about 3.4× as fast as the NVIDIA GeForce RTX 4060 (estimated, median across models).
What is the price difference between the NVIDIA GeForce RTX 3090 and the NVIDIA GeForce RTX 4060?
In the UK the NVIDIA GeForce RTX 3090 costs £1,330 at Amazon and the NVIDIA GeForce RTX 4060 £290 at eBay (used): the NVIDIA GeForce RTX 4060 is £1,040 cheaper.

More comparisons

More hardware for local AI