vram.run Models Hardware Providers Cloud State of Inference
API provider data is live · Hardware & cloud pricing curated 2026-02-23

M4 Max 128GB

Apple Silicon

128GB VRAM · 546 GB/s bandwidth · 36.9 FP16 TFLOPS · 75W TDP · $4,999 street price

The M4 Max 128GB has 128GB of VRAM with 546 GB/s memory bandwidth and 36.9 TFLOPS FP16 compute. At Q4 quantization, it can comfortably run Gemma 3 4B (120 tok/s), Qwen 2.5 7B (63 tok/s), Llama 3.1 8B (59 tok/s). Models larger than ~218B parameters won't fit even at Q4. Electricity cost is approximately $8/month at 75W TDP.

What LLMs can you run?

ModelParamsQ4 WeightFitDecode
Gemma 3 4B4.0B2 GBcomfortable120 tok/s
Qwen 2.5 7B7.6B4 GBcomfortable63 tok/s
Llama 3.1 8B8.0B4 GBcomfortable59 tok/s
Mistral Small 24B24.0B12 GBtight20 tok/s
Gemma 3 27B27.4B14 GBtight17 tok/s
Qwen 2.5 Coder 32B32.5B16 GBtight14 tok/s
Llama 3.3 70B70.6B35 GBtight6 tok/s
Qwen 2.5 72B72.7B36 GBtight6 tok/s
Llama 3.1 405B405B202 GBwon't fit
DeepSeek R1 671B671B336 GBwon't fit

Similar GPUs

GPUVRAMBWTFLOPSTDP
M3 Max 128GB128GB400 GB/s32.875W
M1 Ultra 128GB128GB800 GB/s42.5120W
M2 Ultra 128GB128GB800 GB/s54.4120W
M3 Ultra 128GB128GB800 GB/s65.5120W
GB10128GB273 GB/s29.7140W

Compare with another GPU

Select another GPU to compare specs and model performance side by side.

Install CLI [email protected] Raw data · MIT · API data: live · HW/Cloud data: curated 2026-02-23 · v0.6.0