DiggTop
Menu
⌕

RTX 5090 vs RTX PRO 6000 Blackwell

Pick a different pair →

Best sourced output tok/s per model (single-stream generation) — engine and version, quantization, context and a source link on every row. Hover a value for the quant.

RTX 509023 sourced runs · best 446.7 tok/s
RTX PRO 6000 Blackwell29 verified runs · best 1271.3 tok/s

Specs and evidence

RTX 5090RTX PRO 6000 Blackwell
VRAM32 GB96 GB
Sourced runs on the board2329
Models measured1623
Best output tok/s446.7 tok/s1271.3 tok/s
Highest measured peak VRAM31.3 GB72.7 GB
Street price (latest snapshot)——

Sourced rows only. Peak VRAM is recorded by the person who ran the benchmark, so it is a floor for what the card can hold, not a specification. Price snapshots.

Head-to-head

On the 6 models measured on both cards, RTX 5090 leads on 3, RTX PRO 6000 Blackwell on 3. The widest gap is on Kimi-Linear-48B-A3B-Instruct: RTX 5090 280.8 tok/s vs RTX PRO 6000 Blackwell 90.0 tok/s (RTX 5090 212% faster).

The measured matrix

ModelRTX 5090RTX PRO 6000 Blackwell
MobileLLM-125M—1271.3 Q8_0
Qwen2.5-0.5B-Instruct—1060.0 Q8_0
MobileLLM-350M—980.8 Q8_0
SmolLM2-360M-Instruct—968.1 Q8_0
Qwen3-0.6B—956.4 Q4_K_M
Llama-3.2-1B-Instruct—699.9 Q8_0
gemma-3-1b-it—664.3 Q4_K_M
Llama-3.2-3B-Instruct—489.2 Q4_K_M
NVIDIA-Nemotron-3.5-Lightning-30B-A3B446.7 NVFP4—
Qwen3.8-27B435.1 IQ4_XS296.0 NVFP4
Ling-3.0-tiny—369.1 Q4_K_M
Qwen3.6-35B-A3B—330.2 NVFP4
Ornith-1.0-35B196.6 AWQ309.7 Q4_K_M
Ornith-1.0-35B-PrismaAURA-4.75bit-vllm-MTP283.1 NVFP4—
Kimi-Linear-48B-A3B-Instruct280.8 IQ4_XS90.0 Q8_0
Ornith-1.5-35B-A3B258.8 NVFP4—
North-Mini-Code-1.0258.4 UD-Q6_K_XL—
DeepSeek-Coder-V2-Lite-Instruct—237.3 FP8
Qwen3.6-35B-A3B-MTP222.9 Unsloth-Dynamic-IQ4_XS—
Laguna-XS-2.1216.7 NVFP4—
KAT-Coder-V2.5-Dev-AWQ-W4A16-ASYM201.2 W4A16—
Qwen3-Coder-Next—174.3 FP8
Llama-3.1-8B-Instruct—158.1 FP8
Muse-Glimmer-30B124.8 NVFP4152.2 NVFP4
Qwen3.6-27B120.9 NVFP4101.4 FP8
Huihui-Qwen3.8-27B-abliterated117.7 NVFP4—
Qwen3.8-27B-MTP—112.1 NVFP4
Qwen3.8-27B-5090-goldilocks93.5 Q6_K—
Qwen3-14B—90.8 FP8
Qwen3.6-27B-MTP28.3 Q5_K_M60.1 Q6_K
Qwen2.5-Coder-14B-Instruct—50.3 BF16
DeepSeek-V4-Flash-073126.4 MXFP4—
Qwen3-Coder-30B-A3B-Instruct—26.1 FP8

Head-to-head (models tested on both): RTX 5090 3 · RTX PRO 6000 Blackwell 3 · tied 0 · data as of Oct 17, 2026. Full rules: methodology.

Related reading: RTX 5090 vs 3090 · quantization · Value index · Full board

↑