Local AI Advisor

Can Mac Studio M3 Ultra 512 GB run Nemotron 3.5 Super VL 120B-A12B?

512 GB unified · 819 GB/s memory bandwidth · Nemotron 3.5 Super VL 120B-A12B has 120B parameters (12B active per token).

Short answer

Yes, excellent for chat and comfortable or better for every workload we test.

Best quantization for chat: MXFP4 (60 GB of weights) on vLLM. Generation: 35–45 tok/s, leaving about 440 GB free for your other apps.

How it rates for each workload

Same model and machine, different jobs. Agents and long documents send far bigger prompts than chat, so they need much more speed and memory.

WorkloadVerdictQuantGenerationFirst replyContext
Chatwith Open WebUIExcellentMXFP435–45 tok/s≈1 s8KDetails
Coding questionswith ContinueExcellentMXFP435–45 tok/s≈7 s16KDetails
Coding in a repositorywith AiderComfortableMXFP435–45 tok/s≈20 s32KDetails
Agentic codingwith OpenCodeComfortableMXFP435–45 tok/s≈30 s32KDetails
Long documents (64K)with Open WebUIExcellentMXFP435–45 tok/s≈80 s64KDetails

Speeds are estimates calibrated against measured benchmarks on similar hardware. “First reply” is the time to read the workload’s opening prompt. How we calculate

Different tool, context size or workload?

These pages use typical settings. The full check models your exact setup.