NVIDIA
Operating mode
Use this to bias workload recommendations toward responsiveness, background autonomy, lighter serving, or multi-GPU scale-out.
Current mode
Balanced
Balanced for general local use. Keeps the ranking neutral across personal and serving workflows.
The RTX A6000 48GB is the flagship Ampere workstation GPU and was the go-to professional AI card before Ada Lovelace arrived. Packing 48 GB of ECC GDDR6 at 768 GB/s bandwidth with ISV-certified drivers and NVLink support, it can run 70B quantized models on a single card and scale to 96 GB with a second GPU. While its 768 GB/s bandwidth is 20% lower than the consumer RTX 3090's GDDR6X, the doubled VRAM and enterprise support made it the preferred choice for professional AI inference deployments of its generation.
Beyond LLMs
What AI tasks this GPU can handle — from text generation to image and video creation.
| Capability | Status | Representative Model |
|---|---|---|
| LLM Chat (7B) | Runs natively | Llama 3.1 8B Q4 |
| LLM Coding (30B) | Runs natively | Qwen 3 30B Q4 |
| LLM Large (70B) | Needs offload | Llama 3.1 70B Q4 |
| Image Gen (SDXL) | Runs natively | SDXL 1.0 FP16 |
| Image Gen (Flux) | Runs natively | Flux.1 Dev FP16 |
| Image Gen (SD 3.5) | Runs natively | SD 3.5 Large FP16 |
| Video Short (25f) | Runs natively | LTX Video 2B |
| Video Long (100f) | Won't fit | Wan Video 14B |
Architecture
Ampere is NVIDIA's second-generation RTX architecture, built on Samsung's 8nm process. It introduced 3rd-generation Tensor Cores with support for sparsity-accelerated INT8 operations and improved FP16 throughput over Turing.
AI Relevance
Sparsity-aware Tensor Cores can effectively double throughput for structured sparse workloads. However, the lack of FP8 support means quantized inference is less efficient than Ada Lovelace or Blackwell.
Conselho de compra
Excelente escolha para IA local
Roda 29 de 50 modelos principais bem — um ótimo coringa para inferência local.
48.0 GB
VRAM
$4,650
Preço sugerido
$97/GB
Custo por GB de VRAM
Melhores modelos para esta GPU
What will limit you first
This setup is broadly balanced for this model.
No major red flags
This recommendation has enough memory headroom and acceptable estimated speed for the selected workload.
Best upgrade itinerary
Unlocks 5 additional models that do not fit on the current setup.
Quer mais margem? AMD Instinct MI210 64GB (64.0 GB VRAM) é o próximo passo.
Cost vs cloud API
Assumes 4 hours/day of active inference at 74 tok/s, RTX A6000 48GB amortized over 36 months, US residential electricity ($0.15/kWh), blended cloud pricing at $10 per 1M tokens (GPT-4o / Claude Sonnet tier).
32.1M
Tokens/month at this pace
$139
Monthly local cost
$321
Same tokens on cloud API
$4.33
Local $/1M tokens
Break-even: amortizes in 15.2 months vs cloud API. Price reference: $4.8k MSRP.
Chat
SQwen 3.5 27B matches Chat and keeps a practical fit profile. It is a recent-generation family, which helps on current local SOTA workloads. It fits natively with comfortable headroom. Context coverage stays within the requested workload envelope. Known distribution channels: huggingface, ollama, lm-studio.
Coding
SQwen 3.6 27B is a specialized fit for Coding. It is a recent-generation family, which helps on current local SOTA workloads. It fits natively with comfortable headroom. Context coverage stays within the requested workload envelope. Known distribution channels: huggingface, lm-studio.
Agentic Coding
SQwen 3.6 27B is a specialized fit for Agentic Coding. It is a recent-generation family, which helps on current local SOTA workloads. It fits natively with comfortable headroom. Context coverage stays within the requested workload envelope. Known distribution channels: huggingface, lm-studio.
Reasoning
SDevstral Small 2 24B Instruct matches Reasoning and keeps a practical fit profile. It is a recent-generation family, which helps on current local SOTA workloads. It fits natively with comfortable headroom. Context coverage stays within the requested workload envelope. Known distribution channels: huggingface, ollama, lm-studio.
RAG
SQwen 3.5 27B matches RAG and keeps a practical fit profile. It is a recent-generation family, which helps on current local SOTA workloads. It fits natively with comfortable headroom. Context coverage stays within the requested workload envelope. Known distribution channels: huggingface, ollama, lm-studio.
Qwen 3.6 35B A3BQwen3-Coder 30B A3B InstructQwen 3.5 35B A3BQwen3-VL 30B A3B InstructQwen 3 30B A3BQwen 3.5 27BQwen 3 32BNemotron Cascade 2 30B A3B
Magistral Small 2507Qwen 3.6 27BDevstral Small 2 24B InstructNemotron 3 Nano 30B
GPT-OSS 20B
Gemma 4 31BQwen 3.5 9BQwen 3 14BDevstral Small 1.1
Phi-4-reasoning-plus 14BGemma 4 26B A4BQwen 3 8B
EXAONE 4.0 32BQwen 3.5 4BMinistral 3 14BNemotron Nano 8BPhi-4 Mini Reasoning 4BQwen3-Coder-NextQwen 2.5 VL 72B
Jina Embeddings v3
BGE M3Qwen 3.5 397B A17BDevstral 2 123B Instruct
Kimi K2.5Kimi K2.6
DeepSeek V4 ProQwen 3.5 122B A10BDeepSeek V4 FlashMistral Small 4 119B
Command A 111BGPT-OSS 120B
GLM-5.1Pixtral Large 124BGLM-5DeepSeek V3.2Qwen 3 235B A22BQwen3-Coder 480B A35B InstructLeanstral 119B A6BDeepSeek Coder V2 236BDeepSeek R1 671BDeepSeek V3.1 671BQuase ao alcance
Modelos de alta qualidade que precisam de um pouco mais de memória
Image & Video Generation
50 of 52 models can generate images or video on your RTX A6000 48GB
| Model | Max Resolution | Gen Time | Grade |
|---|---|---|---|
| SD TurboImage | 512×512 | 500ms | S |
| Stable Diffusion 1.5Image | 512×768 | ~1.1s | S |
| Realistic Vision v5.1Image | 512×768 | ~1.1s | S |
| DreamShaper 8Image | 512×768 | ~1.1s | S |
| LCM DreamShaper v7Image | 512×768 | 300ms | S |
| PixArt-SigmaImage | 1024×1024 | ~4.3s | S |
| FramePack I2VVideo | 640×480 | ~13.5s/frame | S |
| SDXL TurboImage | 512×512 | 500ms | S |
| SDXL LightningImage | 1024×1024 | ~1.6s | S |
| Stable Diffusion XL 1.0Image | 1024×1024 | ~4.3s | S |
| Playground v2.5Image | 1024×1024 | ~6.4s | S |
| RealVisXL v5.0Image | 1024×1024 | ~4.8s | S |
| DreamShaper XLImage | 1024×1024 | ~4.8s | S |
| Juggernaut XL v9Image | 1024×1024 | ~4.8s | S |
| Animagine XL 3.1Image | 1024×1024 | ~4.8s | S |
| Pony Diffusion V6 XLImage | 1024×1024 | ~4.8s | S |
| Animagine XL 4.0Image | 1024×1024 | ~4.8s | S |
| Illustrious XLImage | 1024×1024 | ~4.8s | S |
| Wan Video 2.1 1.3BVideo | 480×832 | ~3.1s/frame | S |
| Stable Diffusion 3.5 MediumImage | 1024×1024 | ~7.5s | S |
| Flux.2 Klein 4BImage | 1024×1024 | ~1.3s | S |
| LTX Video 2BVideo | 1280×720 | ~3.7s/frame | S |
| KolorsImage | 1024×1024 | ~8.5s | S |
| Stable CascadeImage | 1024×1024 | ~10.6s | S |
| AuraFlow v0.3Image | 1536×1536 | ~19.2s | S |
| Stable Diffusion 3.5 LargeImage | 1024×1024 | ~23.4s | S |
| Stable Diffusion 3.5 Large TurboImage | 1024×1024 | ~4.3s | S |
| CogVideoX 2BVideo | 720×480 | ~3.7s/frame | S |
| HunyuanVideoVideo | 256×256 | ~13.5s/frame | S |
| ChromaImage | 1024×1024 | ~4.3s | S |
| Z-Image TurboImage | 1536×1536 | ~4.4s | S |
| Flux.1 DevImage | 1024×1024 | ~19.2s | S |
| Flux.1 SchnellImage | 1024×1024 | ~3.7s | S |
| LTX Video 13BVideo | 768×512 | ~7.8s/frame | S |
| Flux.1 Kontext DevImage | 1024×1024 | ~21.3s | S |
| AnimateDiff v1.5.3Video | 512×768 | ~1.9s/frame | S |
| Cosmos Diffusion 7BVideo | 1024×576 | ~6.1s/frame | S |
| CogVideoX 5BVideo | 720×480 | ~5.3s/frame | S |
| Wan2.2 TI2V 5BVideo | 832×480 | ~5.3s/frame | S |
| Flux.2 Klein 9BImage | 1024×1024 | ~2.1s | S |
| Flux.1 Fill DevImage | 1024×1024 | ~18.1s | S |
| Mochi 1 PreviewVideo | 848×480 | ~7s/frame | S |
| HunyuanVideo 1.5Video | 720×1280 | ~6.5s/frame | A |
| Helios 14BVideo | 832×480 | ~8.1s/frame | B |
| SkyReels V2 14BVideo | 256×256 | ~8.1s/frame | B |
| Wan Video 2.1 14BVideo | 256×256 | ~13.8s/frame | D |
| Wan Video 2.2 14BVideo | 256×256 | ~13.8s/frame | D |
| Qwen ImageImage | 256×256 | ~11.8s | D |
| Qwen Image EditImage | 256×256 | ~11.8s | D |
| Flux.2 DevImage | 256×256 | ~3m 22s | D |
| MAGI-1Video | 256×256 | ~10s/frame | F |
| HunyuanImage 3.0Image | 256×256 | ~12.6s | F |
Image models estimated at 1024×1024 (28 steps, FP16). Video models estimated at 768×512 (25 frames, 30 steps, FP16). Actual performance varies with runtime and system load.
Multi-GPU scaling
Scale out with multiple GPUs for larger models. NVLink provides 112.5 GB/s inter-GPU bandwidth with 22% overhead.
| Config | Effective memory | Models that fit | Est. bandwidth |
|---|---|---|---|
| 1× RTX | 48 GB | 338/374 | 768 GB/s |
| 2× RTX | 96 GB | 351/374 | 1,198 GB/s |
Model counts use default quantization at coding workload settings. Multi-GPU scaling factor: 0.78× per additional GPU.
Upgrade paths
See what you unlock with more powerful hardware
Opções de upgrade
2× RTX A6000 48GBMulti-GPUUnlocks 13 additional models that do not fit on the current setup.
Lifts average decode speed across fitting models by about 15%.
Scale-out only pays off if the host platform has enough PCIe lanes, slot spacing, power, and cooling.
The bigger the setup gets, the more the runtime matters. Multi-GPU and multi-user serving are where vLLM, SGLang, TGI, TensorRT-LLM, or tuned llama.cpp start to earn their complexity.
~$4,650 MSRP
Unlocks 5 additional models that do not fit on the current setup.
Lifts average decode speed across fitting models by about 25%.
~$10,000 MSRP
NVIDIA A100 80GBUpgrade NVIDIAUnlocks 12 additional models that do not fit on the current setup.
Lifts average decode speed across fitting models by about 47%.
~$15,000 MSRP
Unlocks 13 additional models that do not fit on the current setup.
~$2,499 MSRP
Unlocks 26 additional models that do not fit on the current setup.
Lifts average decode speed across fitting models by about 135%.
~$8,000 MSRP
RTX A6000 48GB (48 GB VRAM) can run these top models: Qwen 3.6 35B A3B (score: 97/100), Qwen3-Coder 30B A3B Instruct (score: 96/100), Qwen 3.5 35B A3B (score: 96/100). See the full compatibility list above.
RTX A6000 48GB has 48 GB of VRAM available for AI model inference. This determines which models and quantization levels you can run locally.
Yes, RTX A6000 48GB is excellent for running LLMs locally with top compatibility scores above 80/100.
For coding on RTX A6000 48GB, we recommend Qwen 3.6 27B. It achieves 22.7 tokens per second with 262K context window. Qwen 3.6 27B is a specialized fit for Coding. It is a recent-generation family, which helps on current local SOTA workloads. It fits natively with comfortable headroom. Context coverage stays within the requested workload envelope. Known distribution channels: huggingface, lm-studio.
There are 5 upgrade path(s) from RTX A6000 48GB: RTX A6000 48GB, AMD Instinct MI210 64GB. Upgrading would unlock larger models and faster inference speeds.
Yes, RTX A6000 48GB with 48 GB of usable memory can run Flux.1 Dev at FP16 natively. Flux is a 12B parameter diffusion transformer that produces high-quality images. You can also run the Schnell variant for faster generation.
RTX A6000 48GB (48 GB VRAM) can handle various AI generation tasks beyond LLMs. For image generation, SDXL and Stable Diffusion 3.5 run well. Flux.1 Dev also runs natively for state-of-the-art image quality. For video, LTX Video 2.3 can generate short clips. Check the AI Capability Matrix above for detailed compatibility.
RTX A6000 48GB is excellent for AI image generation. With 48 GB of usable memory, it runs all major diffusion models including Flux.1, SDXL, and Stable Diffusion 3.5 at full precision. You can generate high-resolution images quickly and even handle video generation models.
Yes, RTX A6000 48GB with 48 GB of usable memory can run Qwen 3.5 27B at Q8 (near-lossless, ~28.9 GB) or even FP16 (~55.4 GB) depending on your context needs. This setup provides an excellent experience with this model. Use Ollama or vLLM for best results.
With 48 GB VRAM on RTX A6000 48GB, use Q8_0 for most models — it is near-lossless and you have the memory for it. For 70B+ models, Q6_K offers excellent quality. Reserve Q4_K_M for 100B+ models or when you need maximum context length.
RTX A6000 48GB has enough memory for many local LLMs, but bandwidth still matters a lot for real speed. Once a model fits, a faster-memory GPU can feel significantly better than a slower setup with similar capacity.
RTX A6000 48GB supports up to 2× GPU scaling via NVLink at 112.5 GB/s. With 2× GPUs, you get 96 GB effective memory with a 0.78× scaling factor per GPU. This enables running models like Qwen 3.5 397B A17B and Devstral 2 123B Instruct that don't fit on a single card.
NVLink is recommended for RTX A6000 48GB multi-GPU inference, providing 112.5 GB/s interconnect bandwidth with only 22% scaling overhead. PCIe-only setups work but have higher overhead (~25%) due to limited inter-GPU bandwidth.
Compare with similar
