readygpu

NVIDIA GeForce RTX 3090 vs NVIDIA RTX A6000

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 A6000 has more memory for models: 48 GB against 24 GB. At 64k context the NVIDIA RTX A6000 runs 63 of the models we track fully in memory and the NVIDIA GeForce RTX 3090 48; only the NVIDIA RTX A6000 holds Qwen 3.5 122B-A10B. On the models both run fully in memory, the NVIDIA GeForce RTX 3090 is about 1.2× as fast as the NVIDIA RTX A6000 (estimated, median across models).

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

NVIDIA GeForce RTX 3090

2.290 €

Amazon

Add to build
NVIDIA RTX A6000

Not priced here

Add to build

Specifications

SpecNVIDIA GeForce RTX 3090NVIDIA RTX A6000
GPUGeForce RTX 3090RTX A6000
Memory24 GB GDDR6X48 GB GDDR6
Memory bus384-bit384-bit
Memory bandwidth936 GB/s768 GB/s
FP16 compute71 TFLOPS38.71 TFLOPS
SoftwareCUDACUDA
Board power350 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 3090NVIDIA RTX A6000
Qwen 3.8 27B27tok/sQ4_K_M14tok/sQ8_0
Qwen 3.5 122B-A10Boffload56tok/sUD-Q2_K_XL
Qwen 3.6 27B24tok/sQ5_K_M14tok/sQ8_0
Gemma 4 31Boffload12tok/sQ8_0
Mistral Small 4 119Boffload74tok/sUD-Q2_K_XL
Qwen 3.6 35B-A3B107tok/sQ4_K_M64tok/sQ8_0
Qwen 3.5 27B27tok/sQ4_K_M14tok/sQ8_0
Gemma 4 26B-A4B72tok/sUD-Q5_K_M50tok/sQ8_0
Qwen 3.5 35B-A3B107tok/sQ4_K_M64tok/sQ8_0
GLM 4.7 Flash53tok/sQ4_K_M32tok/sQ8_0
Devstral Small 2 24Boffload13tok/sQ8_0
Nemotron 3.5 Lightning 30B-A3B131tok/sMXFP476tok/sUD-Q6_K_XL
gpt-oss 20B86tok/sMXFP471tok/sMXFP4
Kimi Linear 48B-A3Boffload94tok/sQ6_K
Qwen3 Coder 30B-A3Boffload30tok/sQ8_0
Qwen3 32Boffload11tok/sQ6_K
Gemma 4 12B41tok/sQ8_034tok/sQ8_0
Qwen 3.5 9B50tok/sQ8_041tok/sQ8_0
Gemma 3 27B26tok/sQ4_K_M14tok/sQ8_0
Mistral Small 3.2 24Boffload13tok/sQ8_0
DeepSeek R1 Distill Qwen 32Boffload11tok/sQ6_K
Ministral 3 14B27tok/sQ6_K19tok/sQ8_0
Qwen3 14B25tok/sQ6_K18tok/sQ8_0
DeepSeek R1 0528 Qwen3 8B32tok/sQ8_026tok/sQ8_0
Qwen3 8B32tok/sQ8_026tok/sQ8_0
Phi-4 14B24tok/sQ5_K_M16tok/sQ8_0
Qwen 3.5 4B87tok/sQ8_071tok/sQ8_0
Ministral 3 8B32tok/sQ8_026tok/sQ8_0
Granite 4.2 8B29tok/sQ8_024tok/sQ8_0
Gemma 4 E4B57tok/sQ8_047tok/sQ8_0
LFM2.5 8B-A1B81tok/sQ8_067tok/sQ8_0
Llama 3.1 8B38tok/sQ6_K31tok/sQ6_K
Qwen2.5 0.5B434tok/sQ8_0356tok/sQ8_0
Qwen3.5 0.8B355tok/sQ8_0291tok/sQ8_0
Qwen3.5 2B257tok/sQ5_K_M211tok/sQ5_K_M
DeepSeek R1 Distill Qwen 1.5B154tok/sQ8_0126tok/sQ8_0
MiniCPM5 2B106tok/sQ8_087tok/sQ8_0
Spark X2.5 4B105tok/smlx-4bit86tok/smlx-4bit
Mellum2.1 12B A2.5B Thinking96tok/sQ8_078tok/sQ8_0
FrogNano 4B 260985tok/sQ8_070tok/sQ8_0
LFM2.5 2.6B84tok/sQ8_069tok/sQ8_0
LFM2.5 VL 3B84tok/sQ8_069tok/sQ8_0
Kimi VL A3B Thinking 250668tok/sQ8_056tok/sQ8_0
Kimi VL A3B66tok/smlx-8bit54tok/smlx-8bit
granite vision 4.1 4b65tok/sQ8_053tok/sQ8_0
granite 4.2 3b63tok/sQ8_051tok/sQ8_0
Ministral 3 3B Instruct 251255tok/sQ8_045tok/sQ8_0
GLM 4 9B 041453tok/smlx-5bit44tok/smlx-5bit
LFM2.5 24B A2B50tok/sQ5_K_M38tok/sQ8_0
DeepSeek R1 Distill Qwen 7B49tok/sQ8_040tok/sQ8_0
GLM 4.6V Flash46tok/sQ8_038tok/sQ8_0
GLM 4.1V 9B Thinking40tok/smlx-8bit33tok/smlx-8bit
DeepSeek R1 Distill Llama 8B34tok/sQ8_028tok/sQ8_0
Ministral 3 8B Reasoning 251232tok/sQ8_026tok/sQ8_0
Muse Glimmer 30B28tok/sQ5_K_M14tok/sQ8_0
Ministral 3 14B Reasoning 251227tok/sQ6_K19tok/sQ8_0
DeepSeek R1 Distill Qwen 14B25tok/sQ5_K_M17tok/sQ8_0
Devstral Small 2507offload13tok/sQ8_0
Magistral Small 2506offload13tok/sQ8_0
Magistral Small 2507offload13tok/sQ8_0
Magistral Small 2509offload13tok/sQ8_0
Qwen3.8 27B OBLITERATEDoffload19tok/smlx-4bit
granite 4.2 30boffload10tok/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 3090 or the NVIDIA RTX A6000?
Choose the NVIDIA RTX A6000 to run bigger models. Choose the NVIDIA GeForce RTX 3090 for speed.
Which can run bigger models, the NVIDIA GeForce RTX 3090 or the NVIDIA RTX A6000?
At 64k context the NVIDIA RTX A6000 runs 63 of the models we track fully in memory and the NVIDIA GeForce RTX 3090 48; only the NVIDIA RTX A6000 holds Qwen 3.5 122B-A10B.
Which is faster for local AI, the NVIDIA GeForce RTX 3090 or the NVIDIA RTX A6000?
On the models both run fully in memory, the NVIDIA GeForce RTX 3090 is about 1.2× as fast as the NVIDIA RTX A6000 (estimated, median across models).

More comparisons

More hardware for local AI