Apple / unified

Apple M5 Ultra

512 GB decides what fits. 1200 GB/s decides how fast it runs once it does.

Computed for 512 GB, the largest configuration. The Apple M5 Ultra is also sold with 96 or 256 GB, and what fits changes with it.

Run it with 96 GB →Run it with 256 GB →

Open in the calculator →

Find one: Amazon ↗ · eBay (new and used) ↗Store links may pay us a commission. They never decide which card is suggested — the memory arithmetic does.

Quick answer

The Apple M5 Ultra fits 304 of 320 sized models entirely in its 512 GB; the largest widely used one is Hy4-preview (466.2 GB at Q4_K_M). Memory decides what fits; its 1200 GB/s of bandwidth decides how fast it answers.

Largest popular model that fitsHy4-preview780B params · Q4_K_M · needs 466.2 GB~37 tok/s Faster than you readOpen in the calculator →
Most downloaded that fitsGLM-5.3-Flash321B params · Q4_K_M · needs 198.3 GB~71 tok/s Faster than you readOpen in the calculator →

At 8,192 tokens of context with the whole model in device memory. Speeds are estimates from memory bandwidth, not benchmarks run on this card.

Apple M5 Ultra · 512 GB · 1200 GB/s · Q4_K_M where published · 8,192 tokensSpecificationpublished dataDecode speedestimate

304 of 320 fit entirely

304 of the 320 models the engine can size fit entirely in device memory at Q4_K_M where it is published, otherwise the nearest published format, and 8,192 tokens. Unified memory is one pool, so there is no second memory tier to spill into; the remaining 16 do not run at this context.

95%of the models the engine can size fit entirely

512GB of device memory

1200GB/s memory bandwidth

0run with system memory

16do not run at 8,192 tokens

Featured models · Q4_K_M at 8,192 tokens, including offload

Model	Needs	Verdict	Decode	Calculator
Qwen3.5-2B2.3B parameters	2.32 GB	fits	~147 tok/s	Open →
NVIDIA-Nemotron-3-Nano-4B-BF164.0B parameters	3.46 GB	fits	~124 tok/s	Open →
Qwen3-VL-8B-Instruct8.8B parameters	7.04 GB	fits	~92 tok/s	Open →
Qwen3.5-9B9.7B parameters	7.05 GB	fits	~100 tok/s	Open →
gemma-4-12B-it12B parameters	9.54 GB	fits	~77 tok/s	Open →
gpt-oss-20b21B parameters	13.8 GB	fits	~126 tok/s	Open →
Qwen3.8-27B28B parameters	18.6 GB	fits	~53 tok/s	Open →
Qwen3.6-27B28B parameters	18.6 GB	fits	~53 tok/s	Open →
Kimi-Linear-48B-A3B-Instruct49B parameters	30.4 GB	fits	~136 tok/s	Open →
Qwen3.8-Flash-Next180B parameters	111.6 GB	fits	~113 tok/s	Open →
DeepSeek-V4-Flash-Vision-Exp305B parameters	188.1 GB	fits	~64 tok/s	Open →
GLM-5.3-Flash321B parameters	198.3 GB	fits	~71 tok/s	Open →

These examples are selected from prominent labs using the catalogue’s latest Hugging Face 30-day downloads and repository-creation freshness signal, with newer releases guaranteed a place. Offloaded rows assume 32 GB of system RAM, and a speed is only shown for a row that runs. Fitting in memory is not the same as loading: whether the runtime and version you have supports each architecture and format on this machine has not been tested here. Each “Open” link carries the same model, format, context and RAM into the calculator.

Fits entirely in Apple M5 Ultra (M5 Ultra 36-core CPU / 80-core GPU, 512 GB) memory — 304 of 320 sized models

Every model that fitswhole model resident · offload off

Your configuration

The most demanding model that fits is Hy4-preview at Q4_K_M: 466.2 GB of the 512 GB, leaving 45.8 GB spare.

Fully resident at 8,192 tokens (or the model’s own maximum, where that is shorter), offload off, at 1200 GB/s. Each row is a run of the engine for this configuration; the rows start with current, prominent releases and “fits” is memory, not a tested runtime. Older or less prominent models remain available through this search and “Show all”.

Find a model in this list

Showing 40 of 304 models that fit.

Model	Format	Needs	Spare	Decode	Downloads / 30d	Calculator
GLM-5.3-FlashNEWzai-org · 321B params	Q4_K_M	198.3 GB	313.7 GB	~71 tok/s	5.4M	Open →
Qwen3.8-27BQwen · 28B params	Q4_K_M	18.6 GB	493.4 GB	~53 tok/s	6.9M	Open →
DeepSeek-V4-Flash-0731DeepSeek · 304B params	Q4_K_M	187.5 GB	324.5 GB	~64 tok/s	4.5M	Open →
GLM-5.3NEWzai-org · 753B params	Q4_K_M	448.1 GB	63.9 GB	~37 tok/s	1.4M	Open →
Qwen3.8-Flash-NextNEWQwen · 180B params	Q4_K_M	111.6 GB	400.4 GB	~113 tok/s	1.4M	Open →
gemma-4-26B-A4B-itGoogle · 26B params	Q4_K_M	16.8 GB	495.2 GB	~124 tok/s	13M	Open →
DeepSeek-V4-Flash-Vision-ExpNEWDeepSeek · 305B params	Q4_K_M	188.1 GB	323.9 GB	~64 tok/s	915.3K	Open →
gemma-4-31B-itGoogle · 31B params	Q4_K_M	22.2 GB	489.8 GB	~42 tok/s	9.9M	Open →
DeepSeek-V4.1-FlashNEWDeepSeek · 552B params	Q4_K_M	449.4 GB	62.6 GB	~56 tok/s	787.8K	Open →
Qwen3.5-9BQwen · 9.7B params	Q4_K_M	7.05 GB	505.0 GB	~100 tok/s	9M	Open →
Qwen3.5-4BQwen · 4.7B params	Q4_K_M	3.98 GB	508.0 GB	~124 tok/s	7.8M	Open →
Qwen3.6-35B-A3BQwen · 36B params	Q4_K_M	23.1 GB	488.9 GB	~136 tok/s	3.3M	Open →
gemma-4-12B-itGoogle · 12B params	Q4_K_M	9.54 GB	502.5 GB	~77 tok/s	1.9M	Open →
Inkling-Smallthinkingmachines · 266B params	Q4_K_M	165.5 GB	346.5 GB	~83 tok/s	657.4K	Open →
Qwen3.6-27BQwen · 28B params	Q4_K_M	18.6 GB	493.4 GB	~53 tok/s	2.5M	Open →
Qwen3.5-2BQwen · 2.3B params	Q4_K_M	2.32 GB	509.7 GB	~147 tok/s	4.9M	Open →
gemma-4-E4B-itGoogle · 8.0B params	Q4_K_M	5.86 GB	506.1 GB	~119 tok/s	4.4M	Open →
NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16NVIDIA · 32B params	Q4_K_M	20.3 GB	491.7 GB	~43 tok/s	530K	Open →
NVIDIA-Nemotron-3-Nano-4B-BF16NVIDIA · 4.0B params	Q4_K_M	3.46 GB	508.5 GB	~124 tok/s	3.4M	Open →
gemma-4-E2B-itGoogle · 5.1B params	Q4_K_M	4.02 GB	508.0 GB	~149 tok/s	3M	Open →
GLM-5.2zai-org · 753B params	Q4_K_M	448.1 GB	63.9 GB	~37 tok/s	705.8K	Open →
Qwen3.5-0.8BQwen · 873M params	Q4_K_M	1.46 GB	510.5 GB	~162 tok/s	2.6M	Open →
DeepSeek-V4-FlashDeepSeek · 284B params	Q4_K_M	175.9 GB	336.1 GB	~64 tok/s	1.1M	Open →
Qwen3-VL-8B-InstructQwen · 8.8B params	Q4_K_M	7.04 GB	505.0 GB	~92 tok/s	14.6M	Open →
MiniMax-M2.7MiniMaxAI · 229B params	Q4_K_M	141.2 GB	370.8 GB	~75 tok/s	1.1M	Open →
Qwen3.5-27BQwen · 28B params	Q4_K_M	18.6 GB	493.4 GB	~53 tok/s	1.9M	Open →
North-Micro-Vision-InstructCohereLabs · 2.5B params	Q4_K_M	2.99 GB	509.0 GB	~136 tok/s	180.7K	Open →
Hy3Tencent · 299B params	Q4_K_M	181.2 GB	330.8 GB	~55 tok/s	304.9K	Open →
Qwen3.5-35B-A3BQwen · 36B params	Q4_K_M	23.1 GB	488.9 GB	~136 tok/s	1.6M	Open →
NVIDIA-Nemotron-3-Super-120B-A12B-BF16NVIDIA · 124B params	Q4_K_M	76.9 GB	435.1 GB	~14 tok/sAbout reading pace	1.2M	Open →
NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16NVIDIA · 561B params	Q4_K_M	345.4 GB	166.6 GB	~3.2 tok/sSlow	358K	Open →
granite-4.2-8bIBM · 8.8B params	Q4_K_M	7.49 GB	504.5 GB	~87 tok/s	128.4K	Open →
DeepSeek-V3.2DeepSeek · 685B params	Q4_K_M	422.2 GB	89.8 GB	~40 tok/s	3.6M	Open →
GLM-4.7-Flashzai-org · 31B params	Q4_K_M	20.4 GB	491.6 GB	~43 tok/s	1.8M	Open →
granite-4.1-3bIBM · 3.4B params	Q4_K_M	3.72 GB	508.3 GB	~123 tok/s	521.8K	Open →
LFM2.5-2.6BLiquidAI · 2.7B params	Q4_K_M	2.61 GB	509.4 GB	~138 tok/s	108.1K	Open →
MiMo-V2.6-Flash-RLNEWXiaomiMiMo · 309B params	Q4_K_M	192.2 GB	319.8 GB	~70 tok/s	48.6K	Open →
MiniMax-M3MiniMaxAI · 427B params	Q4_K_M	264.0 GB	248.0 GB	~50 tok/s	177.2K	Open →
MiMo-V2.5XiaomiMiMo · 311B params	Q4_K_M	192.2 GB	319.8 GB	~70 tok/s	245.2K	Open →
Qwen3-VL-4B-InstructQwen · 4.4B params	Q4_K_M	4.72 GB	507.3 GB	~111 tok/s	3.5M	Open →

Sized at the format most people actually download, not at FP16. The catalogue ordered by downloads →

Find the Apple M5 Ultra: Amazon ↗ · eBay (new and used) ↗Store links may pay us a commission. They never decide which card is suggested — the memory arithmetic does.

Under the hood

The specification behind every figure

What the manufacturer publishes for this device, and the pages it was read from.

Manufacturer specification

Memory scope	unified-system
Capacity	512 GB
Published options	96 GB, 256 GB, 512 GB
Bandwidth	1200 GB/s
Memory type	unified memory
Bus width	Not published
FP32 peak	Not published
Dense matrix peak	Not published without sparsity
Power	Not published

Source ledger

Mac Studio - Technical Specifications ↗

Retrieved 2026-08-30

Caveats

The flat bandwidth/capacity fields are maxima; configurations preserve Apple's exact CPU/GPU and bandwidth/capacity groupings. No VRAM capacity is claimed.

Official configurations

Configuration	GPU cores	Memory	Bandwidth
M5 Ultra 30-core CPU / 64-core GPU	64	96 GB	1200 GB/s
M5 Ultra 36-core CPU / 80-core GPU	80	256 / 512 GB	1200 GB/s

Published capacity is a hardware ceiling, not guaranteed free runtime memory. The calculator shows the runtime reserve separately rather than folding it into a single number.

Run the diagnostic on the Apple M5 Ultra →