NVIDIA RTX A6000
The NVIDIA RTX A6000 has 48 GB VRAM and 768 GB/s memory bandwidth. It can run 57 of our 97 tracked models natively in VRAM at 8k context.
With 48 GB GDDR6, the NVIDIA RTX A6000 is a workstation-tier GPU that can run 57 models natively. 70B at Q3_K_M native (Q4_K_M's ~51GB total is just past this card's effective capacity). ~25-35 t/s for 7B, ~17 t/s for 70B at Q3_K_M.
The NVIDIA RTX A6000 is an Ampere-generation workstation GPU with 48GB of ECC GDDR6 VRAM. Released in 2020, it bridges consumer RTX cards and data center A100s, running full CUDA drivers, supporting NVLink for multi-GPU setups, and fitting inside standard workstation towers. Its 48GB capacity lets it run 34B models at Q4 and most 70B models with CPU offload, making it a long-running workhorse for on-prem AI labs.
NVIDIA RTX A6000: 2020 Ampere workstation with 48GB GDDR6 at 768 GB/s. Bridges consumer and datacenter.
70B at Q3_K_M native (Q4_K_M's ~51GB total is just past this card's effective capacity). ~25-35 t/s for 7B, ~17 t/s for 70B at Q3_K_M.
Full CUDA with ECC memory. NVLink support for multi-GPU. Long-running workhorse for on-prem AI labs.
| Vendor | NVIDIA |
| Architecture | Ampere |
| VRAM | 48 GB |
| Memory type | GDDR6 |
| Memory bandwidth | 768 GB/s |
| Compute backend | CUDA |
| Tier | Workstation |
| Released | 2020 |
| Models (native) | 57 / 97 |
| Models (offload) | 8 / 97 |
Software: Full llama.cpp and Ollama support out of the box. CUDA 12.x recommended; driver ≥ 525 required.
Cloud GPU Rental
Don't want to buy a NVIDIA RTX A6000? RunPod is a cloud GPU rental service: rent one by the hour instead, no contract, no upfront hardware cost.
Pay by the hour · no contract · pods start in about a minute.
Rent a NVIDIA RTX A6000 on RunPod ↗ (+$5 signup credit)Affiliate link: CanItRun may earn a commission. Doesn't affect the fit calculation above.
Popular models for this GPU
Models this GPU runs natively in VRAM (57)
- Qwen 2.5 72B Instruct72B · MMLU-Pro 71.1Q3_K_M · ~13.4 t/s
- Llama 3.3 70B Instruct70B · MMLU-Pro 68.9Q3_K_M · ~13.7 t/s
- DeepSeek R1 Distill Llama 70B70B · MMLU-Pro 70.0Q3_K_M · ~13.7 t/s
- Llama 3.1 70B Instruct70B · MMLU-Pro 66.4Q3_K_M · ~13.7 t/s
- Mixtral 8x7B Instruct v0.146.7B · MMLU-Pro 29.7Q6_K · ~13.7 t/s
Show 52 more
- Command-R 35B35B · MMLU-Pro 33.0Q6_K · ~12.6 t/s
- Qwen 3.5 35B-A3B (MoE)35B · MMLU-Pro 85.3Q8_0 · ~46.2 t/s
- Qwen 3.6 35B35B · MMLU-Pro 85.2Q8_0 · ~12.7 t/s
- Ornith 1.5 35B-A3B (MoE)35B · MMLU-Pro N/AQ8_0 · ~46.2 t/s
- Yi 1.5 34B Chat34.4B · MMLU-Pro 37.0Q8_0 · ~12.9 t/s
- Qwen3 32B32.8B · MMLU-Pro 65.5Q8_0 · ~13.8 t/s
- Qwen 2.5 32B Instruct32.5B · MMLU-Pro 69.0Q8_0 · ~13.6 t/s
- Qwen 2.5 Coder 32B Instruct32.5B · MMLU-Pro 62.3Q8_0 · ~13.6 t/s
- DeepSeek R1 Distill Qwen 32B32.5B · MMLU-Pro 65.0Q8_0 · ~13.6 t/s
- Nemotron 3 Nano 30B32B · MMLU-Pro 78.3Q8_0 · ~45.1 t/s
- Gemma 4 31B30.7B · MMLU-Pro 85.2Q8_0 · ~14.6 t/s
- Qwen3 30B-A3B (MoE)30B · MMLU-Pro 61.5Q8_0 · ~43.7 t/s
- Nemotron 3.5 Lightning 30B-A3B30B · MMLU-Pro 81.6Q8_0 · ~46.7 t/s
- Muse Glimmer 30B27.8B · MMLU-Pro N/AQ8_0 · ~16.8 t/s
- Gemma 2 27B Instruct27.2B · MMLU-Pro 38.0Q8_0 · ~15.6 t/s
- Gemma 3 27B Instruct27B · MMLU-Pro 67.5Q8_0 · ~16.5 t/s
- Qwen 3.6 27B27B · MMLU-Pro 86.2Q8_0 · ~17.1 t/s
- UI-Mate 27B27B · MMLU-Pro ~86.2Q8_0 · ~17.1 t/s
- Bonsai 27B27B · MMLU-Pro 81.5Ternary (Q2_0) · ~56.7 t/s
- Qwen 3.8 27B27B · MMLU-Pro N/AQ8_0 · ~17.1 t/s
- Gemma 4 26B (MoE)25.2B · MMLU-Pro 82.6Q8_0 · ~36.1 t/s
- Mistral Small 3.1 24B Instruct24B · MMLU-Pro 66.8Q8_0 · ~18.6 t/s
- Mistral Small 22B22.2B · MMLU-Pro 49.2Q8_0 · ~19.6 t/s
- GPT-OSS 20B21B · MMLU-Pro 67.9Q8_0 · ~38.5 t/s
- Qwen3 14B14.8B · MMLU-Pro 61.0BF16 · ~16.1 t/s
- Qwen 2.5 14B Instruct14.7B · MMLU-Pro 63.7BF16 · ~16.1 t/s
- Phi-4 14B Instruct14B · MMLU-Pro 70.4BF16 · ~17 t/s
- Mistral Nemo 12B Instruct12.2B · MMLU-Pro 35.6BF16 · ~19.4 t/s
- Gemma 3 12B Instruct12.2B · MMLU-Pro 60.6BF16 · ~19.6 t/s
- Gemma 4 12B (Unified)12B · MMLU-Pro 77.2BF16 · ~18.3 t/s
- Gemma 2 9B Instruct9.2B · MMLU-Pro 32.0BF16 · ~23.5 t/s
- Qwen 3.5 9B9B · MMLU-Pro 82.5BF16 · ~27.3 t/s
- Ornith 1.5 9B9B · MMLU-Pro N/ABF16 · ~27.3 t/s
- Llama 3.1 8B Instruct8B · MMLU-Pro 48.3BF16 · ~29.2 t/s
- DeepSeek R1 Distill Llama 8B8B · MMLU-Pro 41.0BF16 · ~29.2 t/s
- Qwen3 8B8B · MMLU-Pro 56.7BF16 · ~29 t/s
- Qwen 2.5 7B Instruct7.6B · MMLU-Pro 56.3BF16 · ~31.9 t/s
- Mistral 7B Instruct v0.37.25B · MMLU-Pro 30.0BF16 · ~32.1 t/s
- Gemma 3 4B Instruct4B · MMLU-Pro 43.6BF16 · ~58.7 t/s
- Gemma 4 E4B4B · MMLU-Pro 69.4BF16 · ~55.4 t/s
- Phi-3.5 Mini Instruct3.8B · MMLU-Pro 47.4BF16 · ~46.1 t/s
- Phi-4-mini Instruct3.8B · MMLU-Pro 67.3BF16 · ~57.6 t/s
- Llama 3.2 3B Instruct3.2B · MMLU-Pro 24.0BF16 · ~68 t/s
- Qwen 2.5 3B Instruct3.1B · MMLU-Pro 32.4BF16 · ~76.8 t/s
- Gemma 2 2B Instruct2.6B · MMLU-Pro 17.8BF16 · ~82.2 t/s
- Gemma 4 E2B2B · MMLU-Pro 60.0BF16 · ~113.4 t/s
- SmolLM2 1.7B Instruct1.7B · MMLU-Pro 19.0BF16 · ~99.6 t/s
- Qwen 2.5 1.5B Instruct1.5B · MMLU-Pro 16.8BF16 · ~154.3 t/s
- Llama 3.2 1B Instruct1.24B · MMLU-Pro 12.5BF16 · ~181.6 t/s
- Gemma 3 1B Instruct1B · MMLU-Pro 14.7BF16 · ~214.5 t/s
- Qwen 2.5 0.5B Instruct0.5B · MMLU-Pro 10.0BF16 · ~453.5 t/s
- SmolLM2 360M Instruct0.36B · MMLU-Pro 8.0BF16 · ~472.9 t/s
Models that fit with CPU offload (8)
These use system RAM for layers that don't fit in VRAM, so expect much slower inference.
- Mixtral 8x22B Instruct v0.1141B · MMLU-Pro 40.0Q2_K · ~2.2 t/s
- Mistral Medium 3.5 128B128B · MMLU-Pro N/AQ2_K · ~3 t/s
- Qwen 3.5 122B-A10B (MoE)122B · MMLU-Pro 86.7Q3_K_M · ~5.8 t/s
- Nemotron 3 Super 120B120B · MMLU-Pro 83.7Q3_K_M · ~4.8 t/s
- GPT-OSS 120B117B · MMLU-Pro 80.7Q3_K_M · ~9.5 t/s
- Llama 4 Scout 109B109B · MMLU-Pro 74.3Q3_K_M · ~3.9 t/s
- GLM-4.5 Air 106B106B · MMLU-Pro 81.4Q3_K_M · ~6.9 t/s
- GLM-4.6V 106B106B · MMLU-Pro 79.9Q3_K_M · ~6.9 t/s
Too large for this GPU (32)
- Llama 3.1 405B Instruct
- DeepSeek V3 671B
- DeepSeek R1 671B
- Llama 4 Maverick 400B
- Qwen3 235B-A22B (MoE)
- MiniMax M1 456B
- GLM-4.5 355B
- GLM-4.6 355B
- GLM-4.7 358B
- MiniMax M2.5 229B
- GLM-5 744B
- MiniMax M2.7 229B
- Kimi K2.6
- GLM-5.1 754B
- DeepSeek V4 Pro 1.6T
- DeepSeek V4 Flash 284B
- GLM-5.2 753B
- Nemotron 3 Ultra 550B-A55B
- Step 3.5 Flash
- Step 3.7 Flash
- MiMo V2.5 Pro
- Kimi K2.5
- MiniMax M3
- Inkling
- Kimi K3
- DeepSeek V4 Flash 0731 284B
- Qwen3.8 2.4T-A95B
- Qwen3.8-Flash-Next
- DeepSeek V4 Pro 0813 1.6T
- Ornith 1.5 397B (MoE)
- GLM-5.3 753B
- GLM-5.3-Flash 320B
Compare NVIDIA RTX A6000 with other GPUs
Continue reading
hardware11 min
Best GPU for Running LLMs Locally in 2026vram-guides10 min
Best LLMs for 48 GB VRAM (2026)hardware10 min
Best GPUs for Coding Agents in 2026hardware11 min
Multi-GPU Setups for Local LLMs: The Complete Guidehardware10 min
Best GPUs $1000+ for LLMs (2026)hardware10 min
Used Enterprise GPUs for LLMs: A6000, A100, and BeyondFrequently asked questions
- How much VRAM does the NVIDIA RTX A6000 have?
- The NVIDIA RTX A6000 has 48 GB of GDDR6 with 768 GB/s memory bandwidth.
- What is the NVIDIA RTX A6000 best for?
- With 48 GB of VRAM, the NVIDIA RTX A6000 is ideal for running 70B-class models at Q3-class quantization and large MoE models, a workstation sweet spot for local inference.
- What LLMs can the NVIDIA RTX A6000 run locally?
- The NVIDIA RTX A6000 can run 57 of the 97 open-weight models tracked by CanItRun natively in VRAM at 8k context. Top options include: Qwen 3.8 27B at Q8_0, Ornith 1.5 35B-A3B (MoE) at Q8_0, Ornith 1.5 9B at BF16.
- Can the NVIDIA RTX A6000 run Gemma 4 31B?
- Yes. The NVIDIA RTX A6000 runs Gemma 4 31B natively in VRAM at Q8_0 quantization, achieving approximately 14.6 tokens per second.
- Can the NVIDIA RTX A6000 run Qwen 3.6 27B?
- Yes. The NVIDIA RTX A6000 runs Qwen 3.6 27B natively in VRAM at Q8_0 quantization, achieving approximately 17.1 tokens per second.
- Can the NVIDIA RTX A6000 run Qwen3 8B?
- Yes. The NVIDIA RTX A6000 runs Qwen3 8B natively in VRAM at BF16 quantization, achieving approximately 29 tokens per second.
- Can I rent the NVIDIA RTX A6000 instead of buying it?
- Yes: RunPod and similar cloud GPU providers let you rent NVIDIA RTX A6000 instances by the hour, with no long-term contract. This is often cheaper than buying if you only need it occasionally, and lets you try the GPU before committing to a purchase.