readygpu

NVIDIA GeForce RTX 4090 vs NVIDIA RTX 6000 Ada Generation

For running models locally, priced in Germany. At 64k context the first runs 48 models fully in memory, the second 63.

In short

The NVIDIA RTX 6000 Ada Generation has more memory for models: 48 GB against 24 GB. At 64k context the NVIDIA RTX 6000 Ada Generation runs 63 of the models we track fully in memory and the NVIDIA GeForce RTX 4090 48; only the NVIDIA RTX 6000 Ada Generation holds Qwen 3.5 122B-A10B. On the models both run fully in memory, the NVIDIA GeForce RTX 4090 is about 1.1× as fast as the NVIDIA RTX 6000 Ada Generation (estimated, median across models).

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

NVIDIA GeForce RTX 4090

3.915 €

eBay

Add to build
NVIDIA RTX 6000 Ada Generation

Not priced here

Add to build

Specifications

SpecNVIDIA GeForce RTX 4090NVIDIA RTX 6000 Ada Generation
GPUGeForce RTX 4090RTX 6000 Ada Generation
Memory24 GB GDDR6X48 GB GDDR6
Memory bus384-bit384-bit
Memory bandwidth1,008 GB/s960 GB/s
FP16 compute165 TFLOPS91.1 TFLOPS
SoftwareCUDACUDA
Board power450 W300 W
PCIe generationGen 4Gen 4
PCIe lanesx16x16

What each runs

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

ModelNVIDIA GeForce RTX 4090NVIDIA RTX 6000 Ada Generation
Qwen 3.8 27B29tok/sQ4_K_M18tok/sQ8_0
Qwen 3.5 122B-A10Boffload70tok/sUD-Q2_K_XL
Qwen 3.6 27B26tok/sQ5_K_M18tok/sQ8_0
Gemma 4 31Boffload15tok/sQ8_0
Mistral Small 4 119Boffload92tok/sUD-Q2_K_XL
Qwen 3.6 35B-A3B115tok/sQ4_K_M80tok/sQ8_0
Qwen 3.5 27B30tok/sQ4_K_M18tok/sQ8_0
Gemma 4 26B-A4B78tok/sUD-Q5_K_M63tok/sQ8_0
Qwen 3.5 35B-A3B116tok/sQ4_K_M80tok/sQ8_0
GLM 4.7 Flash57tok/sQ4_K_M41tok/sQ8_0
Devstral Small 2 24Boffload17tok/sQ8_0
Nemotron 3.5 Lightning 30B-A3B141tok/sMXFP495tok/sUD-Q6_K_XL
gpt-oss 20B93tok/sMXFP488tok/sMXFP4
Kimi Linear 48B-A3Boffload117tok/sQ6_K
Qwen3 Coder 30B-A3Boffload37tok/sQ8_0
Qwen3 32Boffload14tok/sQ6_K
Gemma 4 12B44tok/sQ8_042tok/sQ8_0
Qwen 3.5 9B53tok/sQ8_051tok/sQ8_0
Gemma 3 27B28tok/sQ4_K_M17tok/sQ8_0
Mistral Small 3.2 24Boffload17tok/sQ8_0
DeepSeek R1 Distill Qwen 32Boffload14tok/sQ6_K
Ministral 3 14B29tok/sQ6_K24tok/sQ8_0
Qwen3 14B27tok/sQ6_K23tok/sQ8_0
DeepSeek R1 0528 Qwen3 8B34tok/sQ8_032tok/sQ8_0
Qwen3 8B34tok/sQ8_032tok/sQ8_0
Phi-4 14B26tok/sQ5_K_M21tok/sQ8_0
Qwen 3.5 4B94tok/sQ8_089tok/sQ8_0
Ministral 3 8B34tok/sQ8_033tok/sQ8_0
Granite 4.2 8B31tok/sQ8_030tok/sQ8_0
Gemma 4 E4B62tok/sQ8_059tok/sQ8_0
LFM2.5 8B-A1B88tok/sQ8_083tok/sQ8_0
Llama 3.1 8B41tok/sQ6_K39tok/sQ6_K
Qwen2.5 0.5B468tok/sQ8_0445tok/sQ8_0
Qwen3.5 0.8B382tok/sQ8_0364tok/sQ8_0
Qwen3.5 2B277tok/sQ5_K_M263tok/sQ5_K_M
DeepSeek R1 Distill Qwen 1.5B166tok/sQ8_0158tok/sQ8_0
MiniCPM5 2B114tok/sQ8_0108tok/sQ8_0
Spark X2.5 4B113tok/smlx-4bit108tok/smlx-4bit
Mellum2.1 12B A2.5B Thinking103tok/sQ8_098tok/sQ8_0
FrogNano 4B 260992tok/sQ8_087tok/sQ8_0
LFM2.5 2.6B91tok/sQ8_086tok/sQ8_0
LFM2.5 VL 3B91tok/sQ8_086tok/sQ8_0
Kimi VL A3B Thinking 250673tok/sQ8_069tok/sQ8_0
Kimi VL A3B71tok/smlx-8bit67tok/smlx-8bit
granite vision 4.1 4b70tok/sQ8_066tok/sQ8_0
granite 4.2 3b68tok/sQ8_064tok/sQ8_0
Ministral 3 3B Instruct 251259tok/sQ8_056tok/sQ8_0
GLM 4 9B 041457tok/smlx-5bit54tok/smlx-5bit
LFM2.5 24B A2B54tok/sQ5_K_M48tok/sQ8_0
DeepSeek R1 Distill Qwen 7B53tok/sQ8_050tok/sQ8_0
GLM 4.6V Flash49tok/sQ8_047tok/sQ8_0
GLM 4.1V 9B Thinking43tok/smlx-8bit41tok/smlx-8bit
DeepSeek R1 Distill Llama 8B37tok/sQ8_035tok/sQ8_0
Ministral 3 8B Reasoning 251234tok/sQ8_033tok/sQ8_0
Muse Glimmer 30B30tok/sQ5_K_M18tok/sQ8_0
Ministral 3 14B Reasoning 251229tok/sQ6_K24tok/sQ8_0
DeepSeek R1 Distill Qwen 14B27tok/sQ5_K_M21tok/sQ8_0
Devstral Small 2507offload17tok/sQ8_0
Magistral Small 2506offload17tok/sQ8_0
Magistral Small 2507offload17tok/sQ8_0
Magistral Small 2509offload17tok/sQ8_0
Qwen3.8 27B OBLITERATEDoffload24tok/smlx-4bit
granite 4.2 30boffload12tok/sQ8_0
Qwen3 Coder Nextoffloadoffload
Nemotron 3 Super 120B-A12Bwon't runoffload
gpt-oss 120Bwon't runoffload
GLM 4.5 Airwon't runoffload
Qwen3 Next 80B-A3Boffloadoffload
Llama 3.3 70Bwon't runoffload
DeepSeek R1 Distill Llama 70Bwon't runoffload
GLM 4.5Vwon't runoffload
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
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 4.6won'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 4090 or the NVIDIA RTX 6000 Ada Generation?
Choose the NVIDIA RTX 6000 Ada Generation to run bigger models. Choose the NVIDIA GeForce RTX 4090 for speed.
Which can run bigger models, the NVIDIA GeForce RTX 4090 or the NVIDIA RTX 6000 Ada Generation?
At 64k context the NVIDIA RTX 6000 Ada Generation runs 63 of the models we track fully in memory and the NVIDIA GeForce RTX 4090 48; only the NVIDIA RTX 6000 Ada Generation holds Qwen 3.5 122B-A10B.
Which is faster for local AI, the NVIDIA GeForce RTX 4090 or the NVIDIA RTX 6000 Ada Generation?
On the models both run fully in memory, the NVIDIA GeForce RTX 4090 is about 1.1× as fast as the NVIDIA RTX 6000 Ada Generation (estimated, median across models).

More comparisons

More hardware for local AI