NVIDIA A100 80GB SXM4 vs NVIDIA RTX PRO 6000 Blackwell Workstation Edition, AI & Machine Learning Comparison
NVIDIA RTX PRO 6000 Blackwell Workstation Edition wins 144 of 146 benchmarks, averaging 62.1% faster.
Both cards were measured first-party on our bench, same suite, same test rig.
What the numbers say
The gap is widest in Qwen2.5-Coder 32B (Q3_K_M), where NVIDIA RTX PRO 6000 Blackwell Workstation Edition leads by 137% (31.99 vs 75.78 tok/s); the closest fight is Qwen2.5 7B served (4% apart).
Benchmark results head-to-head
| Benchmark | NVIDIA A100 80GB SXM4 | NVIDIA RTX PRO 6000 Blackwell Workstation Edition | Difference |
|---|---|---|---|
| Qwen3 4B tok/s | 197 | 362.81 | -46% |
| Llama 3.1 8B tok/s | 162.57 | 258.37 | -37% |
| Qwen2.5-Coder 14B tok/s | 89.45 | 147.74 | -39% |
| Qwen3 32B tok/s | 45.53 | 70.13 | -35% |
| Llama 3.3 70B tok/s | 24.4 | 34.87 | -30% |
| Stable Diffusion XL images/min | 16.36 | 27.94 | -41% |
| Z-Image Turbo images/min | 9.975 | 15.6 | -36% |
| FLUX.1 dev images/min | 4.264 | 6.921 | -38% |
| FLUX.1 Kontext dev images/min | 1.993 | 3.086 | -35% |
| Qwen-Image-Edit images/min | 1.64 | 2.64 | -38% |
| LTX-Video (distilled) frames/s | 8.95 | 16.36 | -45% |
| DeepSeek-R1 Distill Llama 8B tok/s | 159.37 | 235.07 | -32% |
| DeepSeek-R1 Distill 1.5B tok/s | 327.66 | 635.1 | -48% |
| DeepSeek-R1 Distill 14B tok/s | 87 | 135.41 | -36% |
| DeepSeek-R1 Distill 7B tok/s | 162.96 | 256.04 | -36% |
| Gemma 3 12B tok/s | 91.45 | 138.09 | -34% |
| Gemma 3 4B tok/s | 176.94 | 297.73 | -41% |
| Gemma 4 12B tok/s | 92.43 | 137.73 | -33% |
| Llama 3.2 1B tok/s | 525.32 | 892.74 | -41% |
| Llama 3.2 3B tok/s | 264.25 | 435.63 | -39% |
| Mistral 7B v0.3 tok/s | 171.92 | 247.33 | -30% |
| Mistral Small 24B tok/s | 61.38 | 93.55 | -34% |
| Phi-4 14B tok/s | 95.22 | 142.68 | -33% |
| Phi-4 Mini 3.8B tok/s | 239.17 | 391.04 | -39% |
| Qwen2.5-Coder 7B tok/s | 164.6 | 256.11 | -36% |
| Qwen3 0.6B tok/s | 428.39 | 788.97 | -46% |
| Qwen3 1.7B tok/s | 343.58 | 575.77 | -40% |
| Qwen3 14B tok/s | 90.64 | 139.01 | -35% |
| Qwen3 30B A3B tok/s | 177.32 | 308.74 | -43% |
| Qwen3 8B tok/s | 149.96 | 225.95 | -34% |
| SmolLM3 3B tok/s | 249.76 | 406.45 | -39% |
| Codestral 22B tok/s | 63.26 | 92.7 | -32% |
| DeepSeek-R1 Distill 32B tok/s | 43.11 | 66.06 | -35% |
| Devstral Small 24B tok/s | 61.63 | 93.52 | -34% |
| Dolphin 2.9.1 Yi 1.5 34B tok/s | 43.47 | 63.74 | -32% |
| Dolphin Mistral 24B Venice tok/s | 62.72 | 93.67 | -33% |
| Dolphin X1 8B tok/s | 158.62 | 239.4 | -34% |
| Dolphin 3.0 Llama 3.1 8B tok/s | 158.63 | 239.28 | -34% |
| Dolphin 3.0 R1 Mistral 24B tok/s | 61.17 | 93.62 | -35% |
| Gemma 3 27B tok/s | 48.43 | 72.1 | -33% |
| Qwen2.5-Coder 32B tok/s | 43.36 | 66.07 | -34% |
| Qwen3-Coder 30B A3B tok/s | 182.26 | 317.68 | -43% |
| QwQ 32B tok/s | 43.27 | 66.07 | -35% |
| StarCoder2 15B tok/s | 79.22 | 122.09 | -35% |
| FLUX.1 Schnell images/min | 28.39 | 42.41 | -33% |
| Z-Image Turbo (1024px) images/min | 18.65 | 31.43 | -41% |
| Krea 2 Turbo images/min | 3.35 | 4.71 | -29% |
| BiRefNet images/min | 916.67 | 1445.92 | -37% |
| Depth Anything V2 Large images/min | 876.29 | 1328.11 | -34% |
| Depth Anything V2 Small images/min | 925.15 | 1374.58 | -33% |
| SAM ViT-Base images/min | 744.79 | 1079.89 | -31% |
| SAM ViT-Huge images/min | 145.7 | 193.47 | -25% |
| Swin2SR 4x Upscaler images/min | 28.72 | 52.83 | -46% |
| Qwen2.5 1.5B LoRA train tok/s | 8252.8 | 15753.9 | -48% |
| Qwen2.5 7B LoRA train tok/s | 3900.6 | 6041.8 | -35% |
| SmolLM2 1.7B LoRA train tok/s | 9750.8 | 16255.4 | -40% |
| TinyLlama 1.1B LoRA train tok/s | 9553.3 | 18556.4 | -49% |
| Qwen2.5 1.5B served serve tok/s | 4564.7 | 5786.7 | -21% |
| Qwen2.5 7B served serve tok/s | 2187.2 | 2282.8 | -4% |
| SmolLM2 1.7B served serve tok/s | 4351.4 | 5642 | -23% |
| TinyLlama 1.1B served serve tok/s | 5701.5 | 7678.7 | -26% |
| DeepSeek-R1-Distill-Llama-70B tok/s | 22.99 | 32.52 | -29% |
| DeepSeek-R1-Distill-Qwen-32B-abliterated tok/s | 43.48 | 66.03 | -34% |
| Llama-3.3-70B-Instruct-abliterated tok/s | 22.92 | 32.53 | -30% |
| Meta-Llama-3.1-70B tok/s | 22.92 | 32.52 | -30% |
| Nemotron-3-Nano-30B-A3B tok/s | 204.06 | 325.43 | -37% |
| Qwen2.5-Coder 32B (Q3_K_M) tok/s | 31.99 | 75.78 | -58% |
| AI21-Jamba-Reasoning-3B tok/s | 239.23 | 400.02 | -40% |
| Olmo-3.1-32B-Think tok/s | 46.49 | 65.71 | -29% |
| Codestral 22B (Q3_K_M) tok/s | 49.52 | 107.67 | -54% |
| Dolphin-Mistral-24B-Venice-Edition tok/s | 65.77 | 93.71 | -30% |
| DarkIdol-Llama-3.1-8B-Instruct-1.2-Uncensored tok/s | 164.46 | 239.45 | -31% |
| DeepSeek-Coder-V2-Lite tok/s | 213.91 | 352.52 | -39% |
| DeepSeek-R1-0528-Qwen3-8B tok/s | 152.76 | 226.85 | -33% |
| DeepSeek-R1 Distill 14B (Q3_K_M) tok/s | 67.13 | 149.81 | -55% |
| dolphin-2.9-llama3-8b tok/s | 164.09 | 239.29 | -31% |
| Dolphin X1 Trinity Nano 6B tok/s | 158.41 | 282.02 | -44% |
| EVA-Qwen2.5-14B-v0.2 tok/s | 90.35 | 135.83 | -33% |
| gemma-2-2b-it-abliterated tok/s | 244.41 | 416.68 | -41% |
| gemma-2-2b tok/s | 244.52 | 417.07 | -41% |
| gemma-2-9b tok/s | 108.68 | 156.29 | -30% |
| Gemma 3 12B (Q3_K_M) tok/s | 71.88 | 153.93 | -53% |
| gemma-3-1b tok/s | 324.08 | 567.1 | -43% |
| gemma-3-270m tok/s | 617.58 | 965 | -36% |
| GLM-4.7-Flash-REAP-23B-A3B tok/s | 112.75 | 199.14 | -43% |
| GLM-4.7-Flash tok/s | 124.08 | 215.61 | -42% |
| Josiefied-Qwen3-8B-abliterated-v1 tok/s | 152.29 | 227.15 | -33% |
| gpt-oss-20b tok/s | 212.63 | 375.09 | -43% |
| Hermes-3-Llama-3.2-3B tok/s | 269.13 | 439.65 | -39% |
| Hermes-4-70B tok/s | 24.47 | 32.51 | -25% |
| SmolLM3-3B tok/s | 254.6 | 409.73 | -38% |
| Qwen3-Coder-Next-abliterated tok/s | 119.56 | 210.47 | -43% |
| KAT-Coder-V2.5-Dev tok/s | 146.38 | 253.35 | -42% |
| L3-8B-Stheno-v3.2 tok/s | 164.51 | 239.84 | -31% |
| Laguna-XS-2.1 tok/s | 0 | 0 | n/a |
| LFM2.5-1.2B tok/s | 576.95 | 989.91 | -42% |
| LFM2.5-8B-A1B tok/s | 369.01 | 617.09 | -40% |
| Llama-2-7B tok/s | 180.09 | 263.84 | -32% |
| Llama-3.2-3B-Instruct-uncensored tok/s | 269.13 | 440.15 | -39% |
| Meta-Llama-3.1-8B tok/s | 164.24 | 237.69 | -31% |
| Phi-4-mini tok/s | 243.38 | 393.81 | -38% |
| Mistral-7B-Instruct-v0.1 tok/s | 173.87 | 252.55 | -31% |
| Mistral-7B-Instruct-v0.2 tok/s | 173.8 | 253.34 | -31% |
| Mistral-7B-Instruct-v0.3 tok/s | 173.27 | 253.2 | -32% |
| Mistral-Nemo-Instruct-2407 tok/s | 111.07 | 164.28 | -32% |
| Mistral Small 24B (Q3_K_M) tok/s | 46.63 | 108.88 | -57% |
| Nanbeige4.2-3B tok/s | 0 | 0 | n/a |
| NemoMix-Unleashed-12B tok/s | 110.99 | 164.21 | -32% |
| Hermes-4-14B tok/s | 94.48 | 139.56 | -32% |
| Ornith-1.0-35B tok/s | 138.8 | 237.7 | -42% |
| Ornith-1.0-9B tok/s | 134.67 | 201.69 | -33% |
| phi-2 tok/s | 235.55 | 410.12 | -43% |
| Phi-3.5-mini tok/s | 229.46 | 362.79 | -37% |
| Phi-4 14B (Q3_K_M) tok/s | 82.68 | 165.43 | -50% |
| Qwen-AgentWorld-35B-A3B tok/s | 139.69 | 234.82 | -41% |
| Qwen3-0.6B tok/s | 432.66 | 806.77 | -46% |
| Qwen3-1.7B tok/s | 350.4 | 593.5 | -41% |
| Qwen3-14B tok/s | 94.35 | 139.45 | -32% |
| Qwen3-30B-A3B tok/s | 178.14 | 311.76 | -43% |
| Qwen3-4B-Instruct-2507 tok/s | 199.39 | 333.54 | -40% |
| Qwen3-8B tok/s | 152.8 | 226.88 | -33% |
| Qwen3-Coder-Next tok/s | 118.62 | 205.69 | -42% |
| Qwen3-Next-80B-A3B-Thinking tok/s | 118.93 | 203.29 | -41% |
| Qwen1.5-0.5B tok/s | 540.32 | 1025.58 | -47% |
| Qwen2-1.5B tok/s | 333.79 | 648.49 | -49% |
| Qwen2.5-0.5B tok/s | 571.49 | 905.27 | -37% |
| Qwen2.5-1.5B tok/s | 335.29 | 636.39 | -47% |
| Uncensored tok/s | 89.43 | 135.68 | -34% |
| Qwen2.5-14B tok/s | 90.46 | 135.66 | -33% |
| Qwen2.5-32B tok/s | 45.6 | 66 | -31% |
| Qwen2.5-3B tok/s | 248.47 | 398.72 | -38% |
| Qwen2.5-72B tok/s | 23.95 | 29.65 | -19% |
| Qwen2.5-7B tok/s | 166.92 | 256.63 | -35% |
| Qwen2.5-Coder-0.5B tok/s | 562.59 | 910.13 | -38% |
| Qwen2.5-Coder-1.5B tok/s | 333.87 | 650.4 | -49% |
| Qwen2.5-Coder-14B-Instruct-abliterated tok/s | 90.28 | 135.88 | -34% |
| Qwen2.5-Coder-3B tok/s | 248.7 | 401.24 | -38% |
| Qwen2.5-Coder-7B-Instruct-abliterated tok/s | 166.84 | 257.07 | -35% |
| Qwen3 30B A3B (Q3_K_M) tok/s | 146.72 | 309.61 | -53% |
| Qwen3-4B-Instruct-2507 tok/s | 199.65 | 334.71 | -40% |
| Qwen3-4B-Thinking-2507 tok/s | 199.71 | 334.65 | -40% |
| Qwen3-Coder-Next tok/s | 120.71 | 211.5 | -43% |
| Qwen3-Next-80B-A3B-Thinking tok/s | 120.98 | 214.04 | -43% |
| Qwen3-Next-80B-A3B tok/s | 117.44 | 210.17 | -44% |
| SmolLM2-135M tok/s | 554.44 | 1252.77 | -56% |
| Cydonia-24B-v4.3 tok/s | 65.82 | 93.76 | -30% |
Whole-job comparison
How long each card takes to finish a complete pipeline, not just one model. NVIDIA A100 80GB SXM4 is faster on 0 of 13; NVIDIA RTX PRO 6000 Blackwell Workstation Edition on 13.
| Workflow | NVIDIA A100 80GB SXM4 | NVIDIA RTX PRO 6000 Blackwell Workstation Edition | Difference |
|---|---|---|---|
| 50-image depth pass | 5 s | 4 s | NVIDIA RTX PRO 6000 Blackwell Workstation Edition 1.20x faster |
| 500-image masking run | 3.5 min | 2.6 min | NVIDIA RTX PRO 6000 Blackwell Workstation Edition 1.33x faster |
| 24-frame storyboard | 3.7 min | 3.6 min | NVIDIA RTX PRO 6000 Blackwell Workstation Edition 1.01x faster |
| 60-second AI short film | 4.7 min | 4.5 min | NVIDIA RTX PRO 6000 Blackwell Workstation Edition 1.03x faster |
| 6-panel comic page | 6 min | 5.1 min | NVIDIA RTX PRO 6000 Blackwell Workstation Edition 1.18x faster |
| 200-product catalogue cutout | 7.3 min | 4 min | NVIDIA RTX PRO 6000 Blackwell Workstation Edition 1.81x faster |
| Character sheet, 12 poses | 7.6 min | 6.1 min | NVIDIA RTX PRO 6000 Blackwell Workstation Edition 1.24x faster |
| Full codebase review | 11.2 min | 6.8 min | NVIDIA RTX PRO 6000 Blackwell Workstation Edition 1.65x faster |
| 20 long-form articles | 19.1 min | 13.4 min | NVIDIA RTX PRO 6000 Blackwell Workstation Edition 1.43x faster |
| 40-product photo shoot | 23.5 min | 16.7 min | NVIDIA RTX PRO 6000 Blackwell Workstation Edition 1.41x faster |
| 40-product shoot, start to finish | 25 min | 17.6 min | NVIDIA RTX PRO 6000 Blackwell Workstation Edition 1.42x faster |
| 100-photo restoration batch | 50.8 min | 33.4 min | NVIDIA RTX PRO 6000 Blackwell Workstation Edition 1.52x faster |
| 100-photo restore and enlarge | 54.3 min | 35.3 min | NVIDIA RTX PRO 6000 Blackwell Workstation Edition 1.54x faster |
Specifications compared
| NVIDIA A100 80GB SXM4 | NVIDIA RTX PRO 6000 Blackwell Workstation Edition | |
|---|---|---|
| VRAM | 80GB | 96GB |
| Architecture | Ampere | Blackwell |
| Memory bandwidth | 2039 GB/s | 1792 GB/s |
| Boost clock | 1,410 MHz | 2,617 MHz |
| TDP | 400 W | 600 W |
| Launch MSRP | $17,000 | $8,565 |
| Release | 2020-11-16 | 2025-03-18 |
FAQ
- Which is better for ai & machine learning: NVIDIA A100 80GB SXM4 or NVIDIA RTX PRO 6000 Blackwell Workstation Edition?
- NVIDIA RTX PRO 6000 Blackwell Workstation Edition performs better for ai & machine learning, winning 144 of 146 benchmarks in our suite with an average 62.1% advantage.
- What are the main hardware differences between NVIDIA A100 80GB SXM4 and NVIDIA RTX PRO 6000 Blackwell Workstation Edition?
- NVIDIA A100 80GB SXM4 has 80GB VRAM and a 400W TDP, while NVIDIA RTX PRO 6000 Blackwell Workstation Edition has 96GB VRAM and a 600W TDP.
- Where is the biggest performance difference between NVIDIA A100 80GB SXM4 and NVIDIA RTX PRO 6000 Blackwell Workstation Edition?
- Qwen2.5-Coder 32B (Q3_K_M): NVIDIA RTX PRO 6000 Blackwell Workstation Edition leads by roughly 137% (31.99 vs 75.78 tok/s) in our testing.
NVIDIA A100 80GB SXM4 full review · NVIDIA RTX PRO 6000 Blackwell Workstation Edition full review · All AI & Machine Learning rankings