← AI Hashrate

Can Mac Studio M3 Ultra 96GB run Mistral-Medium-3.5-128B?

Apple · 96 GB LPDDR5 · 819 GB/s bandwidth · Mistral · 128.0B · model ctx up to 256K

Partly — Mistral-Medium-3.5-128B at Q4 fits the Mac Studio M3 Ultra 96GB (96 GB LPDDR5) at 4K context (≈ 84.2 GB) but not at the default 8K (≈ 97.0 GB needed). Use a shorter context or partial offload. Decode at Q4/8K: ≈ 11.4 tok/s (estimated, batch 1). FP16 needs ≈ 282.6 GB — does not fit on this card. Estimates come from memory-bandwidth math; rows tagged measured override estimates. Relative ranking is more reliable than absolute tok/s. Methodology.

Fit & speed by quant and context

QuantContextVRAM neededFits?tok/s (decode)
Q44K84.2 GBYes4.1 est.
Q48K default97.0 GBNo11.4 est.
Q432K173.8 GBNo3.5 est.
FP164K269.8 GBNo0.4 est.
FP168K default282.6 GBNo0.4 est.
FP1632K359.4 GBNo0.2 est.

Fits = weights + KV(ctx) + 1 GB overhead ≤ 95% of VRAM. Measured anchors are context-agnostic; the fit verdict is recomputed per context. A missing tok/s means the model is far beyond this card (offload-only territory).

VRAM breakdown at 8K context

QuantWeightsKV cacheOverheadTotal neededMac Studio M3 Ultra 96GB VRAM
Q470.4 GB25.6 GB1 GB97.0 GB96 GB
FP16256.0 GB25.6 GB1 GB282.6 GB96 GB

Other GPUs that run Mistral-Medium-3.5-128B

All GPUs for Mistral-Medium-3.5-128B →

Other models for the Mac Studio M3 Ultra 96GB

All models on the Mac Studio M3 Ultra 96GB →