NVIDIA A10G vs NVIDIA RTX PRO 6000 Blackwell Workstation Edition, AI & Machine Learning Comparison
NVIDIA RTX PRO 6000 Blackwell Workstation Edition wins 142 of 144 benchmarks, averaging 172.6% faster.
Both cards were measured first-party on our bench, same suite, same test rig.
What the numbers say
The gap is widest in Z-Image Turbo (1024px), where NVIDIA RTX PRO 6000 Blackwell Workstation Edition leads by 397% (6.32 vs 31.43 images/min); the closest fight is Depth Anything V2 Small (46% apart); VRAM decides part of this one: NVIDIA RTX PRO 6000 Blackwell Workstation Edition runs 144 of our 12 AI workloads while the other card runs 137, models that don't fit score zero.
Benchmark results head-to-head
| Benchmark | NVIDIA A10G | NVIDIA RTX PRO 6000 Blackwell Workstation Edition | Difference |
|---|---|---|---|
| Qwen3-4B tok/s | 129.87 | 362.81 | -64% |
| Llama-3.1-8B tok/s | 86.62 | 258.37 | -66% |
| Qwen2.5-Coder-14B tok/s | 47.07 | 147.74 | -68% |
| Qwen3-32B tok/s | 22.04 | 70.13 | -69% |
| Llama 3.3 70B tok/s | 0 | 34.87 | n/a |
| Stable Diffusion XL images/min | 6.38 | 27.94 | -77% |
| Z-Image Turbo images/min | 3.45 | 15.6 | -78% |
| FLUX.1 dev images/min | 0 | 6.921 | n/a |
| FLUX.1 Kontext dev images/min | 0 | 3.086 | n/a |
| Qwen-Image-Edit images/min | 0 | 2.64 | n/a |
| Llama 3.2 1B tok/s | 400.6 | 892.74 | -55% |
| Qwen3 0.6B tok/s | 451.38 | 788.97 | -43% |
| Qwen3 1.7B tok/s | 267.08 | 575.77 | -54% |
| DeepSeek-R1 Distill 1.5B tok/s | 265.75 | 635.1 | -58% |
| Gemma 3 4B tok/s | 124.85 | 297.73 | -58% |
| Llama 3.2 3B tok/s | 170.47 | 435.63 | -61% |
| Mistral 7B v0.3 tok/s | 92.82 | 247.33 | -62% |
| Phi-4 Mini 3.8B tok/s | 144.42 | 391.04 | -63% |
| Qwen2.5-Coder 7B tok/s | 90.05 | 256.11 | -65% |
| SmolLM3 3B tok/s | 172.15 | 406.45 | -58% |
| DeepSeek-R1 Distill Llama 8B tok/s | 86.83 | 235.07 | -63% |
| DeepSeek-R1 Distill 14B tok/s | 46.96 | 135.41 | -65% |
| DeepSeek-R1 Distill 7B tok/s | 89.79 | 256.04 | -65% |
| Gemma 3 12B tok/s | 52.07 | 138.09 | -62% |
| Gemma 4 12B tok/s | 52.67 | 137.73 | -62% |
| Mistral Small 24B tok/s | 30.94 | 93.55 | -67% |
| Phi-4 14B tok/s | 48.18 | 142.68 | -66% |
| Qwen3 14B tok/s | 48.12 | 139.01 | -65% |
| Qwen3 30B A3B tok/s | 140.12 | 308.74 | -55% |
| Qwen3 8B tok/s | 84.09 | 225.95 | -63% |
| Codestral 22B tok/s | 32.09 | 92.7 | -65% |
| DeepSeek-R1 Distill 32B tok/s | 21.85 | 66.06 | -67% |
| Devstral Small 24B tok/s | 30.98 | 93.52 | -67% |
| Dolphin 2.9.1 Yi 1.5 34B tok/s | 21.12 | 63.74 | -67% |
| Dolphin Mistral 24B Venice tok/s | 30.95 | 93.67 | -67% |
| Dolphin X1 8B tok/s | 86.61 | 239.4 | -64% |
| Dolphin 3.0 Llama 3.1 8B tok/s | 86.54 | 239.28 | -64% |
| Dolphin 3.0 R1 Mistral 24B tok/s | 30.9 | 93.62 | -67% |
| Gemma 3 27B tok/s | 24.8 | 72.1 | -66% |
| Qwen2.5-Coder 32B tok/s | 21.92 | 66.07 | -67% |
| Qwen3-Coder 30B A3B tok/s | 142.77 | 317.68 | -55% |
| QwQ 32B tok/s | 21.87 | 66.07 | -67% |
| StarCoder2 15B tok/s | 42.19 | 122.09 | -65% |
| Z-Image Turbo (1024px) images/min | 6.32 | 31.43 | -80% |
| BiRefNet images/min | 432.33 | 1445.92 | -70% |
| Depth Anything V2 Large images/min | 740.74 | 1328.11 | -44% |
| Depth Anything V2 Small images/min | 941.18 | 1374.58 | -32% |
| SAM ViT-Base images/min | 398.72 | 1079.89 | -63% |
| SAM ViT-Huge images/min | 84.22 | 193.47 | -56% |
| Swin2SR 4x Upscaler images/min | 22.31 | 52.83 | -58% |
| Dolphin X1 Trinity Nano 6B tok/s | 171.23 | 282.02 | -39% |
| gpt-oss-20b tok/s | 144.61 | 375.09 | -61% |
| Olmo-3.1-32B-Think tok/s | 22.13 | 65.71 | -66% |
| Dolphin-Mistral-24B-Venice-Edition tok/s | 31 | 93.71 | -67% |
| DeepSeek-Coder-V2-Lite tok/s | 160.13 | 352.52 | -55% |
| DeepSeek-R1-0528-Qwen3-8B tok/s | 83.22 | 226.85 | -63% |
| EVA-Qwen2.5-14B-v0.2 tok/s | 47 | 135.83 | -65% |
| gemma-2-2b-it-abliterated tok/s | 174.56 | 416.68 | -58% |
| gemma-2-2b tok/s | 175.03 | 417.07 | -58% |
| gemma-2-9b tok/s | 57.37 | 156.29 | -63% |
| gemma-3-1b tok/s | 284.78 | 567.1 | -50% |
| gemma-3-270m tok/s | 597.49 | 965 | -38% |
| GLM-4.7-Flash tok/s | 103.24 | 215.61 | -52% |
| SmolLM3-3B tok/s | 171.66 | 409.73 | -58% |
| KAT-Coder-V2.5-Dev tok/s | 113.6 | 253.35 | -55% |
| L3-8B-Stheno-v3.2 tok/s | 86.8 | 239.84 | -64% |
| LFM2.5-1.2B tok/s | 429.96 | 989.91 | -57% |
| Llama-2-7B tok/s | 96.69 | 263.84 | -63% |
| Llama-3.2-3B-Instruct-uncensored tok/s | 167.63 | 440.15 | -62% |
| Meta-Llama-3.1-8B tok/s | 86.69 | 237.69 | -64% |
| Phi-4-mini tok/s | 144.27 | 393.81 | -63% |
| Mistral-7B-Instruct-v0.1 tok/s | 92.45 | 252.55 | -63% |
| Mistral-7B-Instruct-v0.2 tok/s | 92.74 | 253.34 | -63% |
| Mistral-7B-Instruct-v0.3 tok/s | 92.45 | 253.2 | -63% |
| Ornith-1.0-35B tok/s | 104.9 | 237.7 | -56% |
| Ornith-1.0-9B tok/s | 73.78 | 201.69 | -63% |
| Phi-3.5-mini tok/s | 145.97 | 362.79 | -60% |
| Qwen-AgentWorld-35B-A3B tok/s | 104.36 | 234.82 | -56% |
| Qwen3-0.6B tok/s | 446.74 | 806.77 | -45% |
| Qwen3-1.7B tok/s | 262.49 | 593.5 | -56% |
| Qwen3-14B tok/s | 47.8 | 139.45 | -66% |
| Qwen3-8B tok/s | 83.04 | 226.88 | -63% |
| Qwen2.5-0.5B tok/s | 508.78 | 905.27 | -44% |
| Qwen2.5-1.5B tok/s | 265.7 | 636.39 | -58% |
| Qwen2.5-14B tok/s | 47.02 | 135.66 | -65% |
| Qwen2.5-32B tok/s | 21.98 | 66 | -67% |
| Qwen2.5-3B tok/s | 168.23 | 398.72 | -58% |
| Qwen2.5-7B tok/s | 89.68 | 256.63 | -65% |
| Qwen2.5-Coder-1.5B tok/s | 264.79 | 650.4 | -59% |
| Qwen2.5-Coder-3B tok/s | 168 | 401.24 | -58% |
| GLM-4.7-Flash-REAP-23B-A3B tok/s | 97.48 | 199.14 | -51% |
| Josiefied-Qwen3-8B-abliterated-v1 tok/s | 83.26 | 227.15 | -63% |
| LFM2.5-8B-A1B tok/s | 268.04 | 617.09 | -57% |
| Mistral-Nemo-Instruct-2407 tok/s | 57 | 164.28 | -65% |
| Hermes-4-14B tok/s | 47.8 | 139.56 | -66% |
| phi-2 tok/s | 172.45 | 410.12 | -58% |
| Qwen3-30B-A3B tok/s | 140.14 | 311.76 | -55% |
| Qwen2.5-Coder-14B-Instruct-abliterated tok/s | 46.91 | 135.88 | -65% |
| Qwen3-4B-Instruct-2507 tok/s | 129.9 | 334.71 | -61% |
| Qwen3-4B-Thinking-2507 tok/s | 129.99 | 334.65 | -61% |
| SmolLM2-135M tok/s | 680.37 | 1252.77 | -46% |
| Cydonia-24B-v4.3 tok/s | 31.04 | 93.76 | -67% |
| Qwen2.5 1.5B LoRA train tok/s | 4214.4 | 15753.9 | -73% |
| Qwen2.5 7B LoRA train tok/s | 1239 | 6041.8 | -79% |
| SmolLM2 1.7B LoRA train tok/s | 3668.3 | 16255.4 | -77% |
| TinyLlama 1.1B LoRA train tok/s | 5499.6 | 18556.4 | -70% |
| Qwen2.5 1.5B served serve tok/s | 2702.5 | 5786.7 | -53% |
| Qwen2.5 7B served serve tok/s | 858.7 | 2282.8 | -62% |
| SmolLM2 1.7B served serve tok/s | 2242.7 | 5642 | -60% |
| TinyLlama 1.1B served serve tok/s | 3558.2 | 7678.7 | -54% |
| FLUX.1 Schnell images/min | 0 | 42.41 | n/a |
| AI21-Jamba-Reasoning-3B tok/s | 166.29 | 400.02 | -58% |
| Codestral 22B (Q3_K_M) tok/s | 28.01 | 107.67 | -74% |
| DarkIdol-Llama-3.1-8B-Instruct-1.2-Uncensored tok/s | 87.84 | 239.45 | -63% |
| DeepSeek-R1-Distill-Llama-70B tok/s | 0 | 32.52 | n/a |
| DeepSeek-R1 Distill 14B (Q3_K_M) tok/s | 41.72 | 149.81 | -72% |
| DeepSeek-R1-Distill-Qwen-32B-abliterated tok/s | 22.14 | 66.03 | -66% |
| dolphin-2.9-llama3-8b tok/s | 87.98 | 239.29 | -63% |
| Gemma 3 12B (Q3_K_M) tok/s | 46.29 | 153.93 | -70% |
| Hermes-3-Llama-3.2-3B tok/s | 169.23 | 439.65 | -62% |
| Hermes-4-70B tok/s | 0 | 32.51 | n/a |
| Qwen3-Coder-Next-abliterated tok/s | 0 | 210.47 | n/a |
| Laguna-XS-2.1 tok/s | 0 | 0 | n/a |
| Llama-3.3-70B-Instruct-abliterated tok/s | 0 | 32.53 | n/a |
| Meta-Llama-3.1-70B tok/s | 0 | 32.52 | n/a |
| Mistral Small 24B (Q3_K_M) tok/s | 26.8 | 108.88 | -75% |
| Nanbeige4.2-3B tok/s | 0 | 0 | n/a |
| NemoMix-Unleashed-12B tok/s | 57.52 | 164.21 | -65% |
| Nemotron-3-Nano-30B-A3B tok/s | 0 | 325.43 | n/a |
| Phi-4 14B (Q3_K_M) tok/s | 43.56 | 165.43 | -74% |
| Qwen3-4B-Instruct-2507 tok/s | 130.83 | 333.54 | -61% |
| Qwen3-Coder-Next tok/s | 0 | 205.69 | n/a |
| Qwen3-Next-80B-A3B-Thinking tok/s | 0 | 203.29 | n/a |
| Qwen1.5-0.5B tok/s | 542.19 | 1025.58 | -47% |
| Qwen2-1.5B tok/s | 265.45 | 648.49 | -59% |
| Uncensored tok/s | 47.45 | 135.68 | -65% |
| Qwen2.5-72B tok/s | 0 | 29.65 | n/a |
| Qwen2.5-Coder-0.5B tok/s | 513.36 | 910.13 | -44% |
| Qwen2.5-Coder 32B (Q3_K_M) tok/s | 18.78 | 75.78 | -75% |
| Qwen2.5-Coder-7B-Instruct-abliterated tok/s | 91.41 | 257.07 | -64% |
| Qwen3-Coder-Next tok/s | 0 | 211.5 | n/a |
| Qwen3-Next-80B-A3B-Thinking tok/s | 0 | 214.04 | n/a |
| Qwen3-Next-80B-A3B tok/s | 0 | 210.17 | n/a |
| Qwen3 30B A3B (Q3_K_M) tok/s | 120.38 | 309.61 | -61% |
Whole-job comparison
How long each card takes to finish a complete pipeline, not just one model. NVIDIA A10G is faster on 0 of 5; NVIDIA RTX PRO 6000 Blackwell Workstation Edition on 5.
| Workflow | NVIDIA A10G | NVIDIA RTX PRO 6000 Blackwell Workstation Edition | Difference |
|---|---|---|---|
| 50-image depth pass | 7 s | 4 s | NVIDIA RTX PRO 6000 Blackwell Workstation Edition 1.59x faster |
| 500-image masking run | 6 min | 2.6 min | NVIDIA RTX PRO 6000 Blackwell Workstation Edition 2.29x faster |
| 24-frame storyboard | 9 min | 3.6 min | NVIDIA RTX PRO 6000 Blackwell Workstation Edition 2.46x faster |
| 200-product catalogue cutout | 9.6 min | 4 min | NVIDIA RTX PRO 6000 Blackwell Workstation Edition 2.38x faster |
| Full codebase review | 21.3 min | 6.8 min | NVIDIA RTX PRO 6000 Blackwell Workstation Edition 3.15x faster |
Specifications compared
| NVIDIA A10G | NVIDIA RTX PRO 6000 Blackwell Workstation Edition | |
|---|---|---|
| VRAM | 24GB | 96GB |
| Architecture | Ampere | Blackwell |
| Memory bandwidth | 600 GB/s | 1792 GB/s |
| Boost clock | 1,710 MHz | 2,617 MHz |
| TDP | 150 W | 600 W |
| Launch MSRP | $2,800 | $8,565 |
| Release | 2021-11-01 | 2025-03-18 |
FAQ
- Which is better for ai & machine learning: NVIDIA A10G or NVIDIA RTX PRO 6000 Blackwell Workstation Edition?
- NVIDIA RTX PRO 6000 Blackwell Workstation Edition performs better for ai & machine learning, winning 142 of 144 benchmarks in our suite with an average 172.6% advantage.
- What are the main hardware differences between NVIDIA A10G and NVIDIA RTX PRO 6000 Blackwell Workstation Edition?
- NVIDIA A10G has 24GB VRAM and a 150W TDP, while NVIDIA RTX PRO 6000 Blackwell Workstation Edition has 96GB VRAM and a 600W TDP.
- Does VRAM matter more than speed between NVIDIA A10G and NVIDIA RTX PRO 6000 Blackwell Workstation Edition?
- For AI, yes, NVIDIA RTX PRO 6000 Blackwell Workstation Edition fits 21 more of our 12 workloads than NVIDIA A10G. A model that exceeds VRAM doesn't run slower, it doesn't run at all, so the card that fits the model wins that workload outright.
- Where is the biggest performance difference between NVIDIA A10G and NVIDIA RTX PRO 6000 Blackwell Workstation Edition?
- Z-Image Turbo (1024px): NVIDIA RTX PRO 6000 Blackwell Workstation Edition leads by roughly 397% (6.32 vs 31.43 images/min) in our testing.
NVIDIA A10G full review · NVIDIA RTX PRO 6000 Blackwell Workstation Edition full review · All AI & Machine Learning rankings