What LLMs Can the RTX A6000 48GB Run?

The RTX A6000 48GB has 48 GB of VRAM and 768 GB/s of bandwidth. It runs 46 of 58 models fully at Q4_K_M with 8K context, up to Qwen3-Next 80B-A3B, and generates about 90 tokens/s on an 8B model.

NVIDIA · Ampere · Released 2020-10 · Updated October 14, 2026

46 of 58 models fit

RTX A6000 48GB (48.0 GB) at Q4_K_M with 8K context. Largest that fits fully: Qwen3-Next 80B-A3B.

ModelNeedsStatusSpeedBest quant that fits
Qwen 2.5 72B44.5 GBFits~11 tok/sQ4_K_M
Llama 3.3 70B43.2 GBFits~11 tok/sQ4_K_M
DeepSeek R1 Distill Llama 70B43.2 GBFits~11 tok/sQ4_K_M
Mixtral 8x7B28.0 GBFits~50 tok/sQ6_K
Qwen 2.5 32B21.4 GBFits~23 tok/sQ8_0
Qwen 2.5 Coder 32B21.4 GBFits~23 tok/sQ8_0
DeepSeek R1 Distill Qwen 32B21.4 GBFits~23 tok/sQ8_0
Qwen 3 32B21.4 GBFits~23 tok/sQ8_0
Qwen 3.5 35B-A3B21.0 GBFits~126 tok/sQ8_0
Qwen 3.6 35B-A3B21.0 GBFits~126 tok/sQ8_0
Gemma 4 31B19.8 GBFits~25 tok/sQ8_0
Qwen 3 30B-A3B18.8 GBFits~104 tok/sQ8_0
Qwen3-Coder 30B-A3B18.8 GBFits~104 tok/sQ8_0
Gemma 3 27B17.6 GBFits~28 tok/sQ8_0
Qwen 3.5 27B16.9 GBFits~29 tok/sQ8_0
Qwen 3.6 27B16.9 GBFits~29 tok/sQ8_0
Qwen 3.8 27B16.9 GBFits~29 tok/sQ8_0
Mistral Small 3.1 24B15.6 GBFits~32 tok/sQ8_0
Gemma 4 26B-A4B15.6 GBFits~122 tok/sQ8_0
gpt-oss-20b12.9 GBFits~143 tok/sMXFP4 (official)
Qwen 2.5 14B10.7 GBFits~50 tok/sFP16 / BF16
DeepSeek R1 Distill Qwen 14B10.7 GBFits~50 tok/sFP16 / BF16
Phi-4 14B10.6 GBFits~50 tok/sFP16 / BF16
Qwen 3 14B10.5 GBFits~51 tok/sFP16 / BF16
Ministral 3 14B9.7 GBFits~55 tok/sFP16 / BF16
Mistral Nemo 12B9.0 GBFits~60 tok/sFP16 / BF16
Gemma 3 12B8.8 GBFits~61 tok/sFP16 / BF16
Ministral 3 8B6.7 GBFits~85 tok/sFP16 / BF16
Qwen 3 8B6.6 GBFits~87 tok/sFP16 / BF16
Qwen 3.5 9B6.4 GBFits~86 tok/sFP16 / BF16
Llama 3.1 8B6.3 GBFits~90 tok/sFP16 / BF16
DeepSeek R1 Distill Llama 8B6.3 GBFits~90 tok/sFP16 / BF16
Mistral 7B v0.35.7 GBFits~98 tok/sFP16 / BF16
Gemma 4 E4B5.7 GBFits~97 tok/sFP16 / BF16
Qwen 2.5 7B5.6 GBFits~100 tok/sFP16 / BF16
DeepSeek R1 Distill Qwen 7B5.6 GBFits~100 tok/sFP16 / BF16
Qwen 3 4B4.2 GBFits~155 tok/sFP16 / BF16
Phi-4 Mini4.1 GBFits~165 tok/sFP16 / BF16
Gemma 4 E2B4.0 GBFits~151 tok/sFP16 / BF16
Gemma 3 4B3.7 GBFits~167 tok/sFP16 / BF16
Qwen 3.5 4B3.7 GBFits~173 tok/sFP16 / BF16
Ministral 3 3B3.5 GBFits~189 tok/sFP16 / BF16
Llama 3.2 3B3.5 GBFits~195 tok/sFP16 / BF16
Llama 3.2 1B1.7 GBFits~517 tok/sFP16 / BF16
Gemma 3 1B1.6 GBFits~653 tok/sFP16 / BF16
Qwen3-Next 80B-A3B46.3 GBTight fit~114 tok/sIQ4_XS
Qwen 3 235B-A22B135 GB33% on GPU~4.1 tok/s–
MiniMax-M2133 GB34% on GPU~8.2 tok/s–
Qwen 3.5 122B-A10B70.2 GB64% on GPU~15 tok/sQ2_K
Mistral Small 4 119B68.2 GB66% on GPU~24 tok/sQ2_K
gpt-oss-120b67.2 GB67% on GPU~30 tok/sQ2_K
Llama 4 Scout64.0 GB70% on GPU~10 tok/sQ2_K
GLM-4.5-Air62.1 GB73% on GPU~15 tok/sQ2_K
Kimi K2.6566 GBToo small––
DeepSeek R1 671B380 GBToo small––
Llama 3.1 405B234 GBToo small––
Qwen 3.5 397B-A17B226 GBToo small––
DeepSeek V4 Flash161 GBToo small––

RTX A6000 48GB for local AI

A 48 GB workstation card, often found used. One card runs 70B models at Q4_K_M with a modest context. Slower than two RTX 3090s, but one slot, lower power and simpler.

Among the 6 devices we track with about 48 GB of usable memory, the RTX A6000 48GB ranks 3 of 6 for text-generation speed, because generation speed follows memory bandwidth. Others in the same class: RTX 6000 Ada 48GB, L40S 48GB, Apple M4 Max 64GB, Apple M1 Max 64GB. They all hold the same models; the faster ones simply answer sooner.

Our picks for the RTX A6000 48GB

The model we would run first for each job, chosen from our recommendations by what fits fully with an 8K context. "Best quant" is the highest quality that fits; "max context" is at Q4_K_M or the native format.

TaskModelBest quantMax contextSpeed
Best all-rounderLlama 3.3 70BQ4_K_M14K~11 tok/s
CodingQwen 3.6 27BQ8_0256K~29 tok/s
Fastest repliesQwen 3.6 35B-A3BQ8_0256K~126 tok/s
Long documents (64K+)Qwen 3.8 27BQ8_0256K~29 tok/s
Images in, text outGemma 4 31BQ8_0256K~25 tok/s
Image generationFLUX.2 [dev]FP8––
SpecValue
Memory48 GB
Memory bandwidth768 GB/s
ArchitectureAmpere
Released2020-10
Llama 3.1 8B Q4_K_M speed~90 tok/s
Largest model that fitsQwen3-Next 80B-A3B

Image and video models on the RTX A6000 48GB

Peak VRAM with model swapping (ComfyUI default on smaller cards).

ModelFP8FP8 fitGGUF Q4Q4 fit
Stable Diffusion 1.51.8 GBFits1.7 GBFits
SDXL 1.04.1 GBFits4.1 GBFits
Stable Diffusion 3.5 Medium6.4 GBFits6.4 GBFits
Stable Diffusion 3.5 Large8.6 GBFits6.4 GBFits
FLUX.1 [dev]12.1 GBFits7.3 GBFits
FLUX.1 [schnell]12.1 GBFits7.3 GBFits
Qwen-Image20.4 GBFits12.3 GBFits
Wan 2.2 A14B (T2V / I2V)18.8 GBFits13.2 GBFits
Wan 2.2 TI2V 5B8.0 GBFits6.5 GBFits
HunyuanVideo18.3 GBFits13.1 GBFits
FLUX.1 Kontext [dev]12.3 GBFits7.5 GBFits
FLUX.2 [dev]31.0 GBFits23.6 GBFits
Z-Image Turbo6.8 GBFits4.9 GBFits
Qwen-Image-Edit 250920.5 GBFits12.4 GBFits
LTX-219.2 GBFits12.6 GBFits

Upgrade path

The next step up is the RTX PRO 6000 Blackwell 96GB with 96 GB and 133% more bandwidth.

RTX A6000 48GB for LLMs: common questions

Can the RTX A6000 48GB run Llama 3.3 70B?

Yes. Llama 3.3 70B needs about 43.2 GB at Q4_K_M with 8K context, and the best quality that fits is Q4_K_M.

Can the RTX A6000 48GB run Qwen 3 32B?

Yes. Qwen 3 32B needs about 21.4 GB at Q4_K_M with 8K context, and the best quality that fits is Q8_0.

Can the RTX A6000 48GB run Qwen 3 14B?

Yes. Qwen 3 14B needs about 10.5 GB at Q4_K_M with 8K context, and the best quality that fits is FP16 / BF16.

How fast is the RTX A6000 48GB for LLMs?

About 90 tokens per second for an 8B model at Q4_K_M, based on its 768 GB/s memory bandwidth. Larger models scale down roughly in proportion to their size.

Related