Apple - Mac - verified
Local LLMs that run on Mac Studio M5 Ultra (up to 512 GB)
By this directory's rule of thumb, 55 of 63 open-weight models fit in the 512 GB of unified memory on the Mac Studio (M5 Ultra) in its largest configuration. The largest by total parameters is Hy4 preview (780B), whose smallest listed download is 433.4 GB, the 4bit MLX build from mlx-community/Hy4-preview-4bit.
- Models that fit
- 55 of 63
- Memory
- Up to 512 GB
- Memory type
- Unified memory
- Bandwidth
- 1,200 GB/s
Spec sheet
- Memory
- Up to 512 GB
- Configurations
- 96, 256 or 512 GB
- Memory type
- Unified memory
- Unified memory
- Yes, shared by CPU and GPU
- Bandwidth
- 1,200 GB/s
- Released
- 2026
Apple states bandwidth as 1.2TB/s (converted to 1200GB/s), and 512GB requires the M5 Ultra with 36-core CPU and 80-core GPU.
Source: Apple spec page (apple.com)
55 models that fit in 512 GB
Total params, largest first| Model | Total params | Smallest download | Quant | Commercial |
|---|---|---|---|---|
| Hy4 previewTencent | 780B49B active | 433.4 GB | 4bitMLX | Commercial use: Yes |
| GLM-5Z.ai | 753.9B40B active | 418.6 GB | 4bitMLX | Commercial use: Yes |
| GLM-5.2Z.ai | 753.3B | 418.3 GB | 4bitMLX | Commercial use: Yes |
| GLM-5.3Z.ai | 753.3B | 418.3 GB | 4bitMLX | Commercial use: Conditions |
| Mistral Large 3 675B Instruct 2512Mistral AI | 675B41B active | 407.0 GB | Q4_K_MGGUF | Commercial use: Yes |
| MiniMax-M3MiniMax | 427B23B active | 241.5 GB | 4bitMLX | Commercial use: Conditions |
| Llama 4 Maverick 17B-128E InstructMeta | 401.6B17B active | 242.8 GB | Q4_K_MGGUF | Commercial use: Conditions |
| GLM-4.7Z.ai | 358.3B | 198.6 GB | 4bitMLX | Commercial use: Yes |
| GLM-4.6Z.ai | 356.8B | 198.6 GB | 4bitMLX | Commercial use: Yes |
| GLM-5.3-FlashZ.ai | 321.3B18B active | 199.7 GB | UD-Q4_K_XLGGUF | Commercial use: Yes |
| MiMo-V2.6-Flash-RLXiaomi MiMo | 310.8B15B active | 167.4 GB | MXFP4GGUF | Commercial use: Yes |
| Hy3Tencent | 298.8B21B active | 182.3 GB | Q4_K_MGGUF | Commercial use: Yes |
| Step 3.7 FlashStepFun | 201.4B11B active | 111.5 GB | Q4_K_SGGUF | Commercial use: Yes |
| Mistral Medium 3.5 128BMistral AI | 127.7B | 74.9 GB | Q4_K_MGGUF | Commercial use: Conditions |
| Devstral 2 123B Instruct 2512Mistral AI | 125B | 70.3 GB | 4bitMLX | Commercial use: Conditions |
| NVIDIA Nemotron 3.5 Super VL 120B-A12BNVIDIA | 124.5B12B active | 71.8 GB | Q4_0GGUF | Commercial use: Yes |
| Mistral Small 4 119B (2603)Mistral AI | 119.4B6.5B active | 67.8 GB | 4bitMLX | Commercial use: Yes |
| gpt-oss-120bOpenAI | 116.8B5.1B active | 62.3 GB | MXFP4-Q4MLX | Commercial use: Yes |
| Cohere Command A (03-2025)Cohere | 111.1B | 62.5 GB | 4bitMLX | Commercial use: No |
| GLM-4.5-AirZ.ai | 110.5B12B active | 60.1 GB | 4bitMLX | Commercial use: Yes |
| Llama 4 Scout 17B-16E InstructMeta | 108.6B17B active | 61.1 GB | 4bitMLX | Commercial use: Conditions |
| GLM-4.6VZ.ai | 107.7B | 61.9 GB | 4bitMLX | Commercial use: Yes |
| Qwen3-Coder-NextAlibaba Cloud (Qwen) | 79.7B3B active | 44.8 GB | 4bitMLX | Commercial use: Yes |
| NVIDIA Nemotron 3.5 Bolt 74B-A9BNVIDIA | 74.9B9B active | 22.0 GB | 24GB (native 2-bit)GGUF | Commercial use: Yes |
| AI21 Jamba2 MiniAI21 Labs | 51.6B12B active | 31.2 GB | Q4_K_MGGUF | Commercial use: Yes |
| Qwen3.6-35B-A3BAlibaba Cloud (Qwen) | 36B3B active | 20.4 GB | 4bitMLX | Commercial use: Yes |
| Cohere Aya Expanse 32BCohere | 32.3B | 19.8 GB | Q4_K_MGGUF | Commercial use: No |
| Ai2 Olmo 3.1 32B ThinkAi2 | 32.2B | 18.1 GB | 4bitMLX | Commercial use: Yes |
| NVIDIA Nemotron 3.5 Lightning 30B-A3BNVIDIA | 31.6B3B active | 17.8 GB | 4bitMLX | Commercial use: Yes |
| Gemma 4 31BGoogle | 31.3B | 17.7 GB | Q4_0 (QAT)GGUF | Commercial use: Yes |
| IBM Granite 4.2 30BIBM | 29.3B | 16.5 GB | 4bitMLX | Commercial use: Yes |
| Qwen3.8-27BAlibaba Cloud (Qwen) | 27.8B | 16.1 GB | 4bitMLX | Commercial use: Yes |
| Gemma 4 26B A4BGoogle | 25.8B3.8B active | 14.4 GB | Q4_0 (QAT)GGUF | Commercial use: Yes |
| Devstral Small 2 24B Instruct 2512Mistral AI | 24B | 14.3 GB | Q4_K_MGGUF | Commercial use: Yes |
| Liquid AI LFM2.5-24B-A2BLiquid AI | 24B2.5B active | 13.6 GB | 4bitMLX | Commercial use: Conditions |
| gpt-oss-20bOpenAI | 20.9B3.6B active | 11.2 GB | MXFP4-Q4MLX | Commercial use: Yes |
| Phi-4-reasoning-vision-15BMicrosoft | 15.1B | 30.2 GB | BF16Official weights | Commercial use: Yes |
| Phi-4Microsoft | 14.7B | 8.2 GB | 4bitMLX | Commercial use: Yes |
| Phi-4-reasoning-plusMicrosoft | 14.7B | 9.1 GB | Q4_K_MGGUF | Commercial use: Yes |
| Ministral 3 14B Instruct 2512Mistral AI | 13.9B | 8.2 GB | Q4_K_MGGUF | Commercial use: Yes |
| Gemma 4 12BGoogle | 12B | 6.7 GB | 4bitMLX | Commercial use: Yes |
| Qwen3.5-9BAlibaba Cloud (Qwen) | 9.7B | 5.7 GB | Q4_K_MGGUF | Commercial use: Yes |
| Ministral 3 8B Instruct 2512Mistral AI | 8.9B | 5.2 GB | Q4_K_MGGUF | Commercial use: Yes |
| IBM Granite 4.2 8BIBM | 8.8B | 4.9 GB | 4bitMLX | Commercial use: Yes |
| Cohere Command R7B (12-2024)Cohere | 8B | 4.5 GB | 4bitMLX | Commercial use: No |
| Gemma 4 E4BGoogle | 8B | 5.0 GB | Q4_K_MGGUF | Commercial use: Yes |
| Ai2 Olmo Hybrid Instruct DPO 7BAi2 | 7.4B | 14.9 GB | BF16Official weights | Commercial use: Yes |
| Gemma 4 E2BGoogle | 5.1B | 3.1 GB | Q4_K_MGGUF | Commercial use: Yes |
| Qwen3.5-4BAlibaba Cloud (Qwen) | 4.7B | 2.7 GB | Q4_K_MGGUF | Commercial use: Yes |
| Ministral 3 3B Instruct 2512Mistral AI | 3.8B | 2.1 GB | Q4_K_MGGUF | Commercial use: Yes |
| Phi-4-mini-instructMicrosoft | 3.8B | 2.2 GB | 4bitMLX | Commercial use: Yes |
| Phi-4-mini-reasoningMicrosoft | 3.8B | 2.2 GB | 4bitMLX | Commercial use: Yes |
| Liquid AI LFM2.5-2.6BLiquid AI | 2.7B | 1.6 GB | 4bitMLX | Commercial use: Conditions |
| Qwen3.5-2BAlibaba Cloud (Qwen) | 2.3B | 1.3 GB | Q4_K_MGGUF | Commercial use: Yes |
| Qwen3.5-0.8BAlibaba Cloud (Qwen) | 0.9B | 0.5 GB | Q4_K_MGGUF | Commercial use: Yes |
The rest of the 63 need more than 512 GB. Browse every local model.
By memory configuration
Apple sells the Mac Studio (M5 Ultra) with 96, 256 or 512 GB. The same rule applied to each configuration, and the models each step up adds.
| Configuration | Models that fit | Largest that fits | New at this size |
|---|---|---|---|
| 96 GB | 42 / 63 | Mistral Medium 3.5 128B | Base configuration: 42 models |
| 256 GB | 48 / 63 | GLM-4.7 | GLM-4.7, GLM-4.6, GLM-5.3-Flash, MiMo-V2.6-Flash-RL, Hy3, Step 3.7 Flash |
| 512 GB | 55 / 63 | Hy4 preview | Hy4 preview, GLM-5, GLM-5.2, GLM-5.3, Mistral Large 3 675B Instruct 2512, MiniMax-M3, Llama 4 Maverick 17B-128E Instruct |
Compare all 33 GPUs and Macs by memory, bandwidth and how many models fit.
FAQ
How many local LLMs can the Mac Studio (M5 Ultra) run?
55 of the 63 open-weight models in this directory fit in its largest configuration, 512 GB, by this site's rule of thumb that the smallest listed download is at most 85% of the memory. By configuration: 42 at 96 GB, 48 at 256 GB and 55 at 512 GB.
Can the Mac Studio (M5 Ultra) run Llama 4 Maverick 17B-128E Instruct?
Yes, by this directory's rule of thumb: its smallest listed download, 242.8 GB, is at most 85% of 512 GB (435.2 GB), the Q4_K_M GGUF build from unsloth/Llama-4-Maverick-17B-128E-Instruct-GGUF. That leaves room for the context and the operating system. It is a memory check, not a speed estimate.
Can the Mac Studio (M5 Ultra) run NVIDIA Nemotron 3.5 Super VL 120B-A12B?
Yes, by this directory's rule of thumb: its smallest listed download, 71.8 GB, is at most 85% of 512 GB (435.2 GB), the Q4_0 GGUF build from ggml-org/NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-GGUF. That leaves room for the context and the operating system. It is a memory check, not a speed estimate.
Can the Mac Studio (M5 Ultra) run GLM-5.1?
Not by this directory's rule of thumb: its smallest listed download, 464.5 GB, is more than 85% of 512 GB (435.2 GB). It needs more memory than any single device in this list.
Is the Mac Studio (M5 Ultra)'s memory shared with the operating system?
Yes. Apple lists unified memory shared by the CPU and GPU, so the operating system and your apps use the same pool as the model. The 15% headroom in this directory's rule is there for that and for the context.
What memory bandwidth does the Mac Studio (M5 Ultra) have?
Apple states 1,200 GB/s of memory bandwidth (Unified memory). From the spec notes: Apple states bandwidth as 1.2TB/s (converted to 1200GB/s), and 512GB requires the M5 Ultra with 36-core CPU and 80-core GPU. This page does not turn bandwidth into a speed estimate.
Newsletter
New open weights, every week
One email per week. New model releases, open-weight drops and what they mean for what you build.
Keep exploring
Best local AI models
Our curated picks by hardware tier, with benchmarks and the reasoning.
Model API pricing and specs
First-party prices per 1M tokens when you would rather call an API.
AI benchmark leaderboards
Sourced scores, for when you need to compare quality.
Setup guides
Point your coding tools at the model you picked.