readygpu

NVIDIA GeForce RTX 4080 Super vs NVIDIA RTX PRO 4000 Blackwell

For running models locally, priced in the US. At 32k context the first runs 36 models fully in memory, the second 57.

In short

The NVIDIA RTX PRO 4000 Blackwell has more memory for models: 24 GB against 16 GB. At 32k context the NVIDIA RTX PRO 4000 Blackwell runs 57 of the models we track fully in memory and the NVIDIA GeForce RTX 4080 Super 36; only the NVIDIA RTX PRO 4000 Blackwell holds Qwen 3.8 27B. On the models both run fully in memory, the NVIDIA GeForce RTX 4080 Super is about 1.1× as fast as the NVIDIA RTX PRO 4000 Blackwell (estimated, median across models).

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

NVIDIA GeForce RTX 4080 Super

$1,300

Amazon

Add to build
NVIDIA RTX PRO 4000 Blackwell

Not priced here

Add to build

Specifications

SpecNVIDIA GeForce RTX 4080 SuperNVIDIA RTX PRO 4000 Blackwell
GPUGeForce RTX 4080 SuperRTX PRO 4000 Blackwell
Memory16 GB GDDR6X24 GB GDDR7
Memory bus256-bit192-bit
Memory bandwidth736 GB/s672 GB/s
FP16 compute104 TFLOPS46.9 TFLOPS
SoftwareCUDACUDA
Board power320 W145 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 4080 SuperNVIDIA RTX PRO 4000 Blackwell
Qwen 3.8 27Boffload18tok/sQ5_K_M
Qwen 3.6 27Boffload19tok/sQ5_K_M
Gemma 4 31Boffload19tok/sQ4_K_M
Qwen 3.6 35B-A3Boffload96tok/sQ4_K_M
Qwen 3.5 27Boffload19tok/sQ5_K_M
Gemma 4 26B-A4Boffload56tok/sUD-Q6_K
Qwen 3.5 35B-A3Boffload97tok/sQ4_K_M
GLM 4.7 Flashoffload47tok/sQ5_K_M
Devstral Small 2 24Boffload19tok/sQ5_K_M
Nemotron 3.5 Lightning 30B-A3Boffload102tok/sMXFP4
gpt-oss 20B85tok/sMXFP477tok/sMXFP4
Qwen3 Coder 30B-A3Boffload50tok/sQ4_K_M
Gemma 4 12B34tok/sQ8_031tok/sQ8_0
Qwen 3.5 9B43tok/sQ8_039tok/sQ8_0
Gemma 3 27Boffload19tok/sQ5_K_M
Mistral Small 3.2 24Boffload19tok/sQ5_K_M
Qwen3 14B32tok/sQ4_K_M20tok/sQ8_0
Ministral 3 14B30tok/sQ5_K_M21tok/sQ8_0
DeepSeek R1 0528 Qwen3 8B34tok/sQ8_031tok/sQ8_0
Qwen3 8B34tok/sQ8_031tok/sQ8_0
Phi-4 14Boffload19tok/sQ8_0
Qwen 3.5 4B81tok/sQ8_074tok/sQ8_0
Ministral 3 8B34tok/sQ8_031tok/sQ8_0
Granite 4.2 8B31tok/sQ8_028tok/sQ8_0
Gemma 4 E4B50tok/sQ8_045tok/sQ8_0
LFM2.5 8B-A1B102tok/sQ8_094tok/sQ8_0
Llama 3.1 8B42tok/sQ6_K38tok/sQ6_K
Qwen2.5 0.5B489tok/sQ8_0446tok/sQ8_0
Qwen3.5 0.8B370tok/sQ8_0338tok/sQ8_0
Qwen3.5 2B246tok/sQ5_K_M224tok/sQ5_K_M
DeepSeek R1 Distill Qwen 1.5B161tok/sQ8_0147tok/sQ8_0
MiniCPM5 2B112tok/sQ8_0102tok/sQ8_0
Spark X2.5 4B106tok/smlx-4bit97tok/smlx-4bit
LFM2.5 2.6B93tok/sQ8_085tok/sQ8_0
LFM2.5 VL 3B93tok/sQ8_085tok/sQ8_0
Kimi VL A3B Thinking 250693tok/sQ4_K_M61tok/sQ8_0
Mellum2.1 12B A2.5B Thinking86tok/sQ8_079tok/sQ8_0
FrogNano 4B 260979tok/sQ8_073tok/sQ8_0
Kimi VL A3B77tok/smlx-6bit58tok/smlx-8bit
granite vision 4.1 4b72tok/sQ8_066tok/sQ8_0
granite 4.2 3b69tok/sQ8_063tok/sQ8_0
Ministral 3 3B Instruct 251264tok/sQ8_058tok/sQ8_0
GLM 4 9B 041448tok/smlx-5bit43tok/smlx-5bit
DeepSeek R1 Distill Qwen 7B46tok/sQ8_042tok/sQ8_0
GLM 4.6V Flash40tok/sQ8_037tok/sQ8_0
DeepSeek R1 Distill Llama 8B36tok/sQ8_033tok/sQ8_0
GLM 4.1V 9B Thinking35tok/smlx-8bit32tok/smlx-8bit
Ministral 3 8B Reasoning 251234tok/sQ8_031tok/sQ8_0
Ministral 3 14B Reasoning 251230tok/sQ5_K_M21tok/sQ8_0
DeepSeek R1 Distill Qwen 14B30tok/sQ4_K_M19tok/sQ8_0
LFM2.5 24B A2Boffload56tok/sQ6_K
Muse Glimmer 30Boffload17tok/sQ6_K
Devstral Small 2507offload19tok/sQ5_K_M
Magistral Small 2506offload19tok/sQ5_K_M
Magistral Small 2507offload19tok/sQ5_K_M
Magistral Small 2509offload19tok/sQ5_K_M
Qwen3.8 27B OBLITERATEDoffload18tok/smlx-4bit
Qwen 3.5 122B-A10Boffloadoffload
Qwen3 Coder Nextwon't runoffload
Mistral Small 4 119Boffloadoffload
GLM 4.5 Airwon't runoffload
Qwen3 Next 80B-A3Bwon't runoffload
Kimi Linear 48B-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
Nemotron 3 Super 120B-A12Bwon'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 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 4080 Super or the NVIDIA RTX PRO 4000 Blackwell?
Choose the NVIDIA RTX PRO 4000 Blackwell to run bigger models. Choose the NVIDIA GeForce RTX 4080 Super for speed.
Which can run bigger models, the NVIDIA GeForce RTX 4080 Super or the NVIDIA RTX PRO 4000 Blackwell?
At 32k context the NVIDIA RTX PRO 4000 Blackwell runs 57 of the models we track fully in memory and the NVIDIA GeForce RTX 4080 Super 36; only the NVIDIA RTX PRO 4000 Blackwell holds Qwen 3.8 27B.
Which is faster for local AI, the NVIDIA GeForce RTX 4080 Super or the NVIDIA RTX PRO 4000 Blackwell?
On the models both run fully in memory, the NVIDIA GeForce RTX 4080 Super is about 1.1× as fast as the NVIDIA RTX PRO 4000 Blackwell (estimated, median across models).

More comparisons

More hardware for local AI