Mac Studio, M5 Ultra, 512GB

1TB of storage, 1.2TB/s of memory bandwidth, and 504GB of unified memory left for a model once macOS has taken its share. It holds 10 of the 11 models we track.

Offers

checked 10 hours ago

Ships late October 2026. Apple has not published a price.

Fit and Speed below are computed from published specs, so they are already true.

Timing

Collecting price history — 0 of 90 days. No timing verdict yet.

Models it runs — 8K context

Llama-4-Behemoth—
Won't fit— headroom
DeepSeek-V3Q5_K_M
Tight27GB headroomest. 22.9 tok/s
Llama-4-Maverick-17B-128EQ8_0
Fits76GB headroomest. 33.2 tok/s
Qwen3-235B-A22BBF16
Tight31GB headroomest. 13.6 tok/s
Mistral-Large-2411BF16
Fits254GB headroomest. 3.2 tok/s
gpt-oss-120bBF16
Fits268GB headroomest. 58.8 tok/s
Llama-4-Scout-17B-16EBF16
Fits283GB headroomest. 17.6 tok/s
Llama-3.3-70BBF16
Fits360GB headroomest. 5.6 tok/s
Qwen3-32BBF16
Fits437GB headroomest. 12.2 tok/s
Gemma-3-27BBF16
Fits447GB headroomest. 14.4 tok/s
gpt-oss-20bBF16
Fits460GB headroomest. 83.3 tok/s

Model	Best quant	Fit	Headroom	Est. speed
Llama-4-Behemoth	—	Won't fit	—	—
DeepSeek-V3	Q5_K_M	Tight	27GB	22.9 tok/s
Llama-4-Maverick-17B-128E	Q8_0	Fits	76GB	33.2 tok/s
Qwen3-235B-A22B	BF16	Tight	31GB	13.6 tok/s
Mistral-Large-2411	BF16	Fits	254GB	3.2 tok/s
gpt-oss-120b	BF16	Fits	268GB	58.8 tok/s
Llama-4-Scout-17B-16E	BF16	Fits	283GB	17.6 tok/s
Llama-3.3-70B	BF16	Fits	360GB	5.6 tok/s
Qwen3-32B	BF16	Fits	437GB	12.2 tok/s
Gemma-3-27B	BF16	Fits	447GB	14.4 tok/s
gpt-oss-20b	BF16	Fits	460GB	83.3 tok/s

Best quant is the highest-quality one that still loads with headroom. Every speed is an estimate; the formula is here.