AI Models for Mac mini M5 Pro — What runs on 24GB
Specs checked against manufacturer and public documentation on . Results below are estimates, not measurements.
What runs on the Mac mini M5 Pro
All 89 models use the recommended quant when it fits, or the largest smaller tracked fallback for this 24GB configuration. Select any row for the full report and step-by-step setup guide.
| Model | Params | Quant | Needs | Speed | Verdict |
|---|---|---|---|---|---|
| Qwen 3.6 35B-A3B | 35B | IQ4_XSsmaller version | 19.5GB | ~101.2–141.6 tokens/s | ✓ Runs great |
| Qwen 3.6 27B | 27B | Q4_K_M | 18.7GB | ~9.1–12.8 tokens/s | ✓ Runs great |
| Bonsai 27B (1-bit) | 27B | Q1_0 | 5GB | ~40.4–56.6 tokens/s | ✓ Runs great |
| Ternary Bonsai 27B | 27B | Q2_0 | 8.6GB | ~21.3–29.8 tokens/s | ✓ Runs great |
| Gemma 4 26B-A4B | 26B | Q4_0 | 17.7GB | ~69.3–97 tokens/s | ✓ Runs great |
| GPT-OSS 20B | 21B | MXFP4 | 15GB | ~74–103.6 tokens/s | ✓ Runs great |
| Qwen3 14B | 14.8B | Q4_K_M | 11.3GB | ~17.1–23.9 tokens/s | ✓ Runs great |
| Phi-4 14B | 14.7B | Q4_K_M | 11.4GB | ~16.9–23.6 tokens/s | ✓ Runs great |
| Ministral 3 14B | 14B | Q4_K_M | 10.4GB | ~18.7–26.2 tokens/s | ✓ Runs great |
| Gemma 3 12B | 12.2B | Q4_K_M | 9.3GB | ~21–29.4 tokens/s | ✓ Runs great |
| Gemma 4 12B | 12B | Q4_0 | 9GB | ~21.9–30.7 tokens/s | ✓ Runs great |
| Grug 12B | 12B | Q4_K_M | 9.7GB | ~19.9–27.9 tokens/s | ✓ Runs great |
| GRM 3.2 Cliff 9B | 9.4B | Q4_K_M | 7.7GB | ~26–36.4 tokens/s | ✓ Runs great |
| grug 9B (ProCreations) | 9.4B | Q4_K_M | 7.7GB | ~26–36.4 tokens/s | ✓ Runs great |
| Fara 1.5 9B | 9.4B | Q4_K_M | 7.7GB | ~26–36.4 tokens/s | ✓ Runs great |
| Qwen 3.5 9B | 9B | Q4_K_M | 7.4GB | ~26.9–37.7 tokens/s | ✓ Runs great |
| Ornith 1.0 9B | 9B | Q4_K_M | 7.3GB | ~27.4–38.4 tokens/s | ✓ Runs great |
| Qwythos 9B v2 | 9B | Q4_K_M | 7.4GB | ~26.9–37.7 tokens/s | ✓ Runs great |
| UI-Mate 9B | 9B | Q4_K_M | 7.6GB | ~26–36.4 tokens/s | ✓ Runs great |
| Tini Cybersec 8B-A1B | 8.5B | Q4_K_M | 6.9GB | ~250.9–351.3 tokens/s | ✓ Runs great |
| Qwen3 8B | 8.2B | Q4_K_M | 6.6GB | ~30.7–43 tokens/s | ✓ Runs great |
| Llama 3.1 8B | 8B | Q4_K_M | 6.5GB | ~31.3–43.9 tokens/s | ✓ Runs great |
| Ministral 8B | 8B | Q4_K_M | 6.5GB | ~31.3–43.9 tokens/s | ✓ Runs great |
| Ternary Bonsai 8B | 8B | Q2_0 | 3.7GB | ~69.8–97.7 tokens/s | ✓ Runs great |
| Ministral 3 8B | 8B | Q4_K_M | 6.8GB | ~29.5–41.3 tokens/s | ✓ Runs great |
| LFM2.5 8B-A1B | 8B | Q4_K_M | 6.8GB | ~236.2–330.6 tokens/s | ✓ Runs great |
| Granite 4.2 8B | 8B | Q4_K_M | 6.9GB | ~29–40.5 tokens/s | ✓ Runs great |
| Ling 3.0 Tiny | 7.9B | Q4_K_M | 6.5GB | ~176.8–247.5 tokens/s | ✓ Runs great |
| DeepSeek R1 Distill 7B | 7.6B | Q4_K_M | 6.3GB | ~32.7–45.7 tokens/s | ✓ Runs great |
| Mistral 7B v0.3 | 7.2B | Q4_K_M | 5.9GB | ~34.9–48.8 tokens/s | ✓ Runs great |
| AREX Turbo 4B | 4.5B | Q4_K_M | 4.2GB | ~52.9–74.1 tokens/s | ✓ Runs great |
| Fara 1.5 4B | 4.5B | Q4_K_M | 4.2GB | ~52.9–74.1 tokens/s | ✓ Runs great |
| Gemma 3 4B | 4.3B | Q4_K_M | 3.7GB | ~61.4–86 tokens/s | ✓ Runs great |
| Nanbeige 4.2 3B | 4.2B | Q4_K_M | 3.9GB | ~56.9–79.6 tokens/s | ✓ Runs great |
| Qwen3 4B | 4B | Q4_K_M | 3.7GB | ~61.4–86 tokens/s | ✓ Runs great |
| Qwen 3.5 4B | 4B | Q4_K_M | 3.9GB | ~56.9–79.6 tokens/s | ✓ Runs great |
| Gemma 4 E4B | 4B | Q4_K_M | 6.3GB | ~30.7–43 tokens/s | ✓ Runs great |
| Ternary Bonsai 4B | 4B | Q2_0 | 2.2GB | ~139.5–195.4 tokens/s | ✓ Runs great |
| Nemotron 3 Nano 4B | 4B | Q4_K_M | 4GB | ~54.8–76.8 tokens/s | ✓ Runs great |
| Agents-A1 4B | 4B | Q4_K_M | 3.9GB | ~56.9–79.6 tokens/s | ✓ Runs great |
| Phi-4 Mini 3.8B | 3.8B | Q4_K_M | 3.7GB | ~61.4–86 tokens/s | ✓ Runs great |
| Llama 3.2 3B | 3.2B | Q4_K_M | 3.1GB | ~76.8–107.5 tokens/s | ✓ Runs great |
| SmolLM3 3B | 3.1B | Q4_K_M | 3GB | ~80.8–113.1 tokens/s | ✓ Runs great |
| LFM2.5-VL 3B | 3.1B | Q4_K_M | 2.8GB | ~90.3–126.4 tokens/s | ✓ Runs great |
| Ministral 3 3B | 3B | Q4_K_M | 3.2GB | ~73.1–102.3 tokens/s | ✓ Runs great |
| G9v3 3B | 3B | Q4_K_M | 3GB | ~80.8–113.1 tokens/s | ✓ Runs great |
| Granite 4.2 3B | 3B | Q4_K_M | 3.3GB | ~69.8–97.7 tokens/s | ✓ Runs great |
| LFM2.5 2.6B | 2.7B | Q4_K_M | 2.8GB | ~90.3–126.4 tokens/s | ✓ Runs great |
| Qwen 3.5 2B | 2B | Q4_K_M | 2.3GB | ~118.1–165.3 tokens/s | ✓ Runs great |
| Gemma 4 E2B | 2B | Q4_K_M | 4.2GB | ~49.5–69.3 tokens/s | ✓ Runs great |
| DeepSeek R1 Distill 1.5B | 1.8B | Q4_K_M | 2.1GB | ~139.5–195.4 tokens/s | ✓ Runs great |
| Qwen3 1.7B | 1.7B | Q8_0 | 2.8GB | ~85.3–119.4 tokens/s | ✓ Runs great |
| SmolLM2 1.7B | 1.7B | Q4_K_M | 2.1GB | ~139.5–195.4 tokens/s | ✓ Runs great |
| Ternary Bonsai 1.7B | 1.7B | Q2_0 | 1.4GB | ~307–429.8 tokens/s | ✓ Runs great |
| Llama 3.2 1B | 1.2B | Q4_K_M | 1.7GB | ~191.9–268.6 tokens/s | ✓ Runs great |
| Gemma 3 1B | 1B | Q4_K_M | 1.7GB | ~191.9–268.6 tokens/s | ✓ Runs great |
| OvisOCR2 0.8B | 0.8B | Q4_K_M | 1.5GB | ~255.8–358.2 tokens/s | ✓ Runs great |
| Qwen3 0.6B | 0.6B | Q8_0 | 1.5GB | ~255.8–358.2 tokens/s | ✓ Runs great |
| BigBang V1 36B-A3B | 36B | Q3_K_Msmaller version | 20.9GB | ~110.3–154.4 tokens/s | ! Runs, barely |
| Qwen 3.5 35B-A3B | 35B | Q3_K_Msmaller version | 20.5GB | ~109.2–152.9 tokens/s | ! Runs, barely |
| XYZ-Aquila mini 35B-A3B | 35B | Q3_K_Msmaller version | 20.3GB | ~110.5–154.8 tokens/s | ! Runs, barely |
| KAT-Coder V2.5 Dev | 35B | Q3_K_Msmaller version | 20.3GB | ~110.5–154.8 tokens/s | ! Runs, barely |
| Gemma 4 31B | 31B | IQ4_XSsmaller version | 20.2GB | ~9.4–13.1 tokens/s | ! Runs, barely |
| Nemotron 3.5 Lightning 30B-A3B | 30B | Q4_0smaller version | 20.7GB | ~81.2–113.7 tokens/s | ! Runs, barely |
| Salience 1.5 Flash | 30B | IQ4_XSsmaller version | 20.2GB | ~84.6–118.4 tokens/s | ! Runs, barely |
| Granite 4.2 30B | 30B | Q4_K_Ssmaller version | 20.4GB | ~9.2–12.9 tokens/s | ! Runs, barely |
| Muse Glimmer 30B | 29.6B | K_QUANT_17GB | 20.5GB | ~9.1–12.8 tokens/s | ! Runs, barely |
| Qwen 3.5 27B | 27B | Q4_K_M | 20.2GB | ~9.2–12.9 tokens/s | ! Runs, barely |
| Fara 1.5 27B | 27B | Q4_K_Ssmaller version | 20GB | ~9.3–13 tokens/s | ! Runs, barely |
| Qwen3.8 27B | 27B | Q4_K_M | 21GB | ~8.1–11.3 tokens/s | ! Runs, barely |
| UI-Mate 27B | 27B | Q4_K_Ssmaller version | 20GB | ~9.3–13 tokens/s | ! Runs, barely |
| Qwen3.8 2.4T-A95B | 2400B | IQ4_XS | 1377.6GB | — | ✕ Won't fit |
| Inkling | 952.4B | Q8_0 | 967.1GB | — | ✕ Won't fit |
| LongCat Flash Chat | 561.9B | IQ1_S | 159.8GB | — | ✕ Won't fit |
| Ornith 1.0 397B | 397B | Q4_K_M | 285.8GB | — | ✕ Won't fit |
| Hunyuan 3 (Hy3) | 298.8B | Q4_K_M | 213GB | — | ✕ Won't fit |
| DeepSeek V4 Flash 0731 | 284B | Q4_K_XL | 163.8GB | — | ✕ Won't fit |
| Inkling Small | 276B | Q4_K_M | 190.7GB | — | ✕ Won't fit |
| Ling 3.0 Flash | 124B | Q4_K_M | 82.5GB | — | ✕ Won't fit |
| Laguna S 2.1 | 118B | Q4_K_M | 109.9GB | — | ✕ Won't fit |
| GLM-4.5V | 106B | Q4_K_M | 75GB | — | ✕ Won't fit |
| GLM-4.5-Air | 106B | Q4_K_M | 75GB | — | ✕ Won't fit |
| Llama 3.3 70B | 70B | Q4_K_M | 50.3GB | — | ✕ Won't fit |
| Ornith 1.0 35B-A3B | 35B | Q4_K_M | 25.5GB | — | ✕ Won't fit |
| Laguna XS 2.1 | 33B | Q4_K_M | 24.4GB | — | ✕ Won't fit |
| Qwen3 32B | 32.8B | Q4_K_M | 23.9GB | — | ✕ Won't fit |
| Nemotron 3 Nano Omni 30B-A3B | 31B | Q4_K_M | 26.5GB | — | ✕ Won't fit |
| Qwen3 30B A3B | 30.5B | Q4_K_M | 22.5GB | — | ✕ Won't fit |
| Nemotron 3 Nano 30B-A3B | 30B | Q4_K_M | 28.7GB | — | ✕ Won't fit |
~ = bandwidth-based estimate · open a row to see the recommended app and exact steps
Other Mac mini M5 Pro memory options
FAQ
What is the biggest local AI model the Mac mini M5 Pro · 24GB can run?
BigBang V1 36B-A3B is the largest model in our 89-model comparison that fits at Q3_K_M. It needs about 20.9GB inside our 21GB working-memory budget, with an estimated 110.3–154.4 tokens/s decode range.
How much of the 24GB unified memory is available to a local LLM?
We use a conservative 21GB working budget, leaving room for macOS, the inference app, and normal background activity. Memory pressure, context length, and other open apps can change the real limit.
Can the Mac mini M5 Pro · 24GB run Llama 3.1 8B?
Yes. At Q4_K_M, our estimate uses 6.5GB and lands around 31.3–43.9 tokens/s.
Can the Mac mini M5 Pro · 24GB run Qwen 3.6 27B?
Yes at Q4_K_M: about 18.7GB of working memory and an estimated 9.1–12.8 tokens/s.
Can the Mac mini M5 Pro · 24GB run DeepSeek V4 Flash 0731?
No tracked quant fits our conservative 21GB working-memory budget. The recommended Q4_K_XL alone needs about 163.8GB at 4K context.
Which app should I use for local AI on the Mac mini M5 Pro?
For most exact GGUF repositories and quants on this site, start with LM Studio's graphical Discover, download, load, and chat flow. Jan is the open-source GGUF alternative; Ollama is strongest when the exact model already has a trustworthy catalog package or you want coding/API integrations; Msty is useful for mixed GGUF, MLX, and document workflows. DeepSeek V4 Flash uses a separate Unsloth Desktop path because of its sharded GGUF packaging. For the curated Bonsai 27B build, try Locally AI and first confirm that its catalog shows the model on this Mac.
Can I upgrade the Mac mini M5 Pro to more unified memory later?
No. Apple-silicon unified memory is integrated into the chip package and must be chosen at purchase. Some Mac mini owners use “upgrade” to mean internal or external storage, but storage capacity does not give an LLM more RAM.
Are these Mac mini M5 Pro local AI speeds measured?
No. Every speed on this page is a formula range based on Apple’s published memory bandwidth, model size, active parameters, and a cooling-aware efficiency range. App, backend, context length, and thermals can change real results.