Mac M3 Max 128G — what LLMs can it run?
100 of 110 indexed models fit comfortably in 128GB at 4K context, each at the highest-quality quant that still leaves headroom.
The short answer for Mac M3 Max 128G
128 GB, 400 GB/s Unified LPDDR5. 100 of 110 models in this index fit comfortably at 4K context; 100 load at all.
- Biggest that fits
- DBRX Instruct — 84.3 GB at Q4_K_M — 11.7 GB spare at 4K, so longer context comes out of a thin margin.
- Room to grow
- Llama 3.2 90B Vision Instruct — 57.5 GB at Q4_K_M — under 60% of the card, which leaves 38.5 GB for a long context window or a second process.
- Speed ceiling
- Llama 3.1 8B Instruct at Q4_K_M reads 4.6 GB of weights per token, so 400 GB/s puts a hard ceiling near 87 tok/s. That is arithmetic on two published numbers, not a benchmark — real throughput lands below it. No run on this card has been measured here, so there is nothing to compare the ceiling against.
- The wall you will hit
- The GPU's share of unified memory, not the 128 GB on the box. macOS reserves part of the pool for the system and caps what a single process may wire down, so the practical budget is meaningfully below nameplate — the limit is adjustable (`iogpu.wired_limit_mb`) but it is not absent. Bandwidth is 400 GB/s, which is the number that decides tok/s once a model fits.
70B+ · 13
| Model | Quant | Est. VRAM | Headroom | tok/s |
|---|---|---|---|---|
| DBRX Instruct132B MoE | Q4_K_M | 84.31 GB | +11.7 GB | ≤19Ceiling |
| Qwen3.8-Flash-Next125B-A6B | Q4_K_M | 79.32 GB | +16.7 GB | ≤116Ceiling |
| Nemotron 3.5 Super VL 120B-A12B120B-A12B | Q4_0 | 75.12 GB | +20.9 GB | ≤59Ceiling |
| GPT-OSS 120B117B MoE | MXFP4 | 66.4 GB | +29.6 GB | ≤152Ceiling |
| Llama 4 Scout 17B (16E)109B MoE | Q4_K_M | 69.88 GB | +26.1 GB | ≤41Ceiling |
| GLM-4.5-Air106B MoE | Q4_K_M | 67.94 GB | +28.1 GB | ≤58Ceiling |
| Llama 3.2 90B Vision Instruct90B | Q4_K_M | 57.5 GB | +38.5 GB | ≤8Ceiling |
| Nemotron 3.5 Bolt 74B-A9B74B-A9B | IQ2_XXS | 23.05 GB | +73 GB | ≤157Ceiling |
| Qwen2.5 72B Instruct72B | Q5_K_M | 55.31 GB | +40.7 GB | ≤8Ceiling |
| Llama 3.1 70B Instruct70B | Q5_K_M | 53.75 GB | +42.3 GB | ≤8Ceiling |
| Llama 3.3 70B Instruct70B | Q5_K_M | 53.75 GB | +42.3 GB | ≤8Ceiling |
| DeepSeek-R1-Distill-Llama-70B70B | Q4_K_M | 46.1 GB | +49.9 GB | ≤10Ceiling |
| Jamba 1.5 Mini52B-A12B | Q4_K_M | 33.01 GB | +63 GB | ≤58Ceiling |
32B · 30
| Model | Quant | Est. VRAM | Headroom | tok/s |
|---|---|---|---|---|
| Kimi Linear 48B-A3B Instruct48B-A3B | Q8_0 | 53.33 GB | +42.7 GB | ≤132Ceiling |
| Mixtral 8x7B Instruct47B MoE | Q4_K_M | 30.13 GB | +65.9 GB | ≤54Ceiling |
| Seed-OSS 36B Instruct36B | Q5_K_M | 27.81 GB | +68.2 GB | ≤16Ceiling |
| Qwen3.6 35B-A3B35B-A3B | Q8_0 | 39.6 GB | +56.4 GB | ≤132Ceiling |
| Command R 35B35B | Q4_K_M | 22.86 GB | +73.1 GB | ≤20Ceiling |
| Yi 1.5 34B Chat34B | Q4_K_M | 22.82 GB | +73.2 GB | ≤20Ceiling |
| Qwen3 32B Instruct32B | Q4_K_M | 21.85 GB | +74.2 GB | 13–17Derived |
| Qwen2.5 32B Instruct32B | Q4_K_M | 21.69 GB | +74.3 GB | 13–17Derived |
| Qwen2.5-Coder 32B Instruct32B | Q4_K_M | 21.69 GB | +74.3 GB | 13–17Derived |
| DeepSeek-R1-Distill-Qwen-32B32B | Q4_K_M | 21.69 GB | +74.3 GB | 13–17Derived |
| Nemotron 3.5 Lightning 30B-A3B30B-A3B | Q8_0 | 35.14 GB | +60.9 GB | ≤132Ceiling |
| Gemma 4 31B IT31B | Q8_0 | 35.72 GB | +60.3 GB | 8–10Derived |
| Qwen3 30B-A3B Instruct30B-A3B | Q5_K_M | 23.04 GB | +73 GB | ≤198Ceiling |
| Qwen3-Coder 30B-A3B Instruct30B-A3B | Q5_K_M | 23.04 GB | +73 GB | ≤198Ceiling |
| Qwen3-VL 30B-A3B Instruct30B-A3B | Q8_0 | 34.28 GB | +61.7 GB | ≤132Ceiling |
| GLM-4.7-Flash30B-A3B | Q8_0 | 33.47 GB | +62.5 GB | ≤132Ceiling |
| Granite 4.2 30B30B | Q8_0 | 33.61 GB | +62.4 GB | 8–11Derived |
| Muse Glimmer 30B30B | Q8_0 | 31.08 GB | +64.9 GB | 8–11Derived |
| Qwen3.8 27B27B | Q4_K_M | 17.47 GB | +78.5 GB | 15–20Derived |
| Qwen3.6 27B27B | Q8_0 | 30.71 GB | +65.3 GB | 9–11Derived |
| Gemma 3 27B IT27B | Q4_K_M | 18.37 GB | +77.6 GB | 15–20Derived |
| Gemma 2 27B Instruct27B | Q5_K_M | 21.76 GB | +74.2 GB | 13–17Derived |
| Gemma 4 26B-A4B IT26B-A4B | Q8_0 | 28.42 GB | +67.6 GB | ≤99Ceiling |
| Mistral Small 24B Instruct24B | Q4_K_M | 15.89 GB | +80.1 GB | 17–23Derived |
| LFM2.5 24B-A2B24B-A2B | Q8_0 | 26.73 GB | +69.3 GB | ≤162Ceiling |
| Devstral Small 1.1 24B24B | Q6_K | 20.91 GB | +75.1 GB | 13–17Derived |
| Magistral Small 1.2 24B24B | Q6_K | 20.91 GB | +75.1 GB | 13–17Derived |
| Codestral 22B22B | Q4_K_M | 15.03 GB | +81 GB | 19–25Derived |
| ERNIE 4.5 21B-A3B21B-A3B | Q8_0 | 24.44 GB | +71.6 GB | ≤132Ceiling |
| GPT-OSS 20B21B MoE | MXFP4 | 12.76 GB | +83.2 GB | ≤202Ceiling |
14B · 18
| Model | Quant | Est. VRAM | Headroom | tok/s |
|---|---|---|---|---|
| InternLM2 20B Chat20B | Q5_K_M | 15.59 GB | +80.4 GB | 18–24Derived |
| DeepSeek-Coder-V2-Lite Instruct16B-A2.4B | Q8_0 | 17.56 GB | +78.4 GB | ≤165Ceiling |
| DeepSeek-V2-Lite Chat16B-A2.4B | Q4_K_M | 10.08 GB | +85.9 GB | ≤289Ceiling |
| StarCoder2 15B15B | Q4_K_M | 10.19 GB | +85.8 GB | 28–37Derived |
| Qwen3 14B Instruct14B | Q5_K_M | 11.65 GB | +84.4 GB | 24–32Derived |
| Qwen2.5 14B Instruct14B | Q5_K_M | 11.73 GB | +84.3 GB | 24–32Derived |
| DeepSeek-R1-Distill-Qwen-14B14B | Q4_K_M | 10.14 GB | +85.9 GB | 28–37Derived |
| Phi-4 14B14B | Q5_K_M | 11.77 GB | +84.2 GB | 24–32Derived |
| Phi-3 Medium 14B Instruct14B | Q6_K | 12.86 GB | +83.1 GB | 22–29Derived |
| Mistral Nemo 12B Instruct12B | Q6_K | 11.14 GB | +84.9 GB | 25–33Derived |
| Gemma 3 12B IT12B | Q5_K_M | 9.84 GB | +86.2 GB | 29–38Derived |
| Mellum2.1 12B-A2.5B Thinking12B-A2.5B | Q8_0 | 13.6 GB | +82.4 GB | ≤159Ceiling |
| Stable LM 2 12B Chat12B | Q4_K_M | 8.35 GB | +87.7 GB | 34–45Derived |
| Llama 3.2 11B Vision Instruct11B | Q8_0 | 12.9 GB | +83.1 GB | 21–28Derived |
| Solar 10.7B Instruct11B | Q4_K_M | 7.6 GB | +88.4 GB | 39–51Derived |
| Falcon 3 10B Instruct10B | Q4_K_M | 7.28 GB | +88.7 GB | 40–53Derived |
| Gemma 2 9B Instruct9B | Q8_0 | 11.7 GB | +84.3 GB | 26–34Derived |
| Qwen3.5 9B9B | Q8_0 | 10.37 GB | +85.6 GB | 26–34Derived |
7B · 25
| Model | Quant | Est. VRAM | Headroom | tok/s |
|---|---|---|---|---|
| GLM-4-9B-Chat9B | Q8_0 | 10.16 GB | +85.8 GB | 26–35Derived |
| Qwen3-VL 8B Instruct8B | Q8_0 | 10.39 GB | +85.6 GB | 27–36Derived |
| Granite 4.2 8B8B | Q8_0 | 10.45 GB | +85.6 GB | 27–36Derived |
| Ministral 3 8B Instruct8B | Q8_0 | 10.02 GB | +86 GB | 28–37Derived |
| Qwen2-VL 7B Instruct7B | Q4_K_M | 5.49 GB | +90.5 GB | 50–66Derived |
| Granite 3.1 8B Instruct8B | Q4_K_M | 5.88 GB | +90.1 GB | 51–67Derived |
| Qwen3 8B Instruct8B | Q6_K | 7.64 GB | +88.4 GB | 37–49Derived |
| Llama 3.1 8B Instruct8B | Q8_0 | 9.47 GB | +86.5 GB | 29–39Derived |
| Nous Hermes 3 Llama 3.1 8B8B | Q4_K_M | 5.64 GB | +90.4 GB | 52–68Derived |
| Aya 23 8B8B | Q4_K_M | 5.64 GB | +90.4 GB | 52–68Derived |
| OpenChat 3.6 8B8B | Q4_K_M | 5.64 GB | +90.4 GB | 52–68Derived |
| DeepSeek-R1-Distill-Llama-8B8B | Q5_K_M | 6.51 GB | +89.5 GB | 44–58Derived |
| InternLM2 7B Chat7B | Q4_K_M | 5.45 GB | +90.6 GB | 53–71Derived |
| Qwen2.5 7B Instruct7B | Q6_K | 6.77 GB | +89.2 GB | 40–53Derived |
| Qwen2.5-Coder 7B Instruct7B | Q4_K_M | 5.07 GB | +90.9 GB | 54–72Derived |
| DeepSeek-R1-Distill-Qwen-7B7B | Q4_K_M | 5.07 GB | +90.9 GB | 54–72Derived |
| Gemma 4 E4B ITE4B | Q8_0 | 8.46 GB | +87.5 GB | 31–42Derived |
| OLMo 2 7B Instruct7B | Q8_0 | 10.3 GB | +85.7 GB | 32–43Derived |
| OLMo 3 7B Instruct7B | Q8_0 | 10.3 GB | +85.7 GB | 32–43Derived |
| WizardLM-2 7B7B | Q4_K_M | 5.14 GB | +90.9 GB | 57–76Derived |
| Mistral 7B Instruct v0.37B | Q6_K | 6.75 GB | +89.3 GB | 42–56Derived |
| Zephyr 7B Beta7B | Q6_K | 6.75 GB | +89.3 GB | 42–56Derived |
| Gemma 4 E2B ITE2B | Q8_0 | 5.2 GB | +90.8 GB | ≤85Ceiling |
| Qwen3.5 4B4B | Q8_0 | 4.97 GB | +91 GB | ≤91Ceiling |
| Gemma 3 4B IT4B | Q8_0 | 5.05 GB | +91 GB | ≤92Ceiling |
≤3B · 14
| Model | Quant | Est. VRAM | Headroom | tok/s |
|---|---|---|---|---|
| Qwen3 4B Instruct4B | Q6_K | 4.05 GB | +92 GB | ≤128Ceiling |
| Phi-4 Mini Instruct3.8B | Q8_0 | 4.81 GB | +91.2 GB | ≤103Ceiling |
| Phi-3.5 Mini Instruct3.8B | Q8_0 | 5.89 GB | +90.1 GB | ≤104Ceiling |
| Granite 4.2 3B3B | Q8_0 | 4.41 GB | +91.6 GB | ≤108Ceiling |
| Llama 3.2 3B Instruct3B | Q8_0 | 4.05 GB | +92 GB | ≤123Ceiling |
| Qwen2.5 3B Instruct3B | Q8_0 | 3.59 GB | +92.4 GB | ≤128Ceiling |
| SmolLM3 3B3B | Q8_0 | 3.73 GB | +92.3 GB | ≤129Ceiling |
| Gemma 2 2B Instruct2B | Q8_0 | 3.34 GB | +92.7 GB | ≤152Ceiling |
| Qwen3.5 2B2B | Q8_0 | 2.23 GB | +93.8 GB | ≤202Ceiling |
| Qwen3 1.7B Instruct1.7B | Q8_0 | 2.39 GB | +93.6 GB | ≤230Ceiling |
| Qwen2.5 1.5B Instruct1.5B | Q8_0 | 1.83 GB | +94.2 GB | ≤258Ceiling |
| Llama 3.2 1B Instruct1B | Q8_0 | 1.51 GB | +94.5 GB | ≤320Ceiling |
| Qwen3.5 0.8B0.8B | Q8_0 | 0.94 GB | +95.1 GB | ≤494Ceiling |
| Qwen2.5 0.5B Instruct0.5B | Q8_0 | 0.6 GB | +95.4 GB | ≤816Ceiling |
How this list is built
Each row is the lowest-perplexity-loss quant of that model whose estimated total — weights plus KV cache at 4K context plus activation buffer — uses at most 88% of the card. That is the calculator's "green" threshold, so every row here has real headroom rather than only just fitting. Raise the context length and the list shortens; the calculator lets you check any combination directly. The speed column is this card: a measured run, a range scaled from this card's own Llama 2 7B benchmark, or a bandwidth ceiling. A figure that only exists for the RTX 4090 is labelled as that card.
Measured on this card
No benchmark runs in this index were recorded on a Mac M3 Max 128G. Every figure on this page is calculated from the model architecture and the quant level — treat them as estimates, not measurements.
Cards with the same budget
What fits is decided by memory, so every 128GB card of this type returns the same list. These pages are not different answers — they differ in throughput, which this index does not measure per card.
The same 128GB in system memory is a different machine — everything on this page fits there, at a speed set by your DIMMs rather than a graphics bus: 128 GB RAM (CPU)
Same list, different speed — play one sentence on both cards: Mac M3 Max 128G vs Mac M4 Max 128G (Qwen3 32B Instruct)
Setup guides for this card
Step-by-step guides written for the Mac M3 Max 128G's runtime or memory size, with the commands and the checks that prove the model is running on the card.
Stepping up
A Mac M3 Ultra 192G (192GB) fits 1 more of the indexed models than this card. Mac M3 Ultra 192G →
Before buying, hear the speed difference on a model both cards hold: Mac M3 Max 128G vs Mac M3 Ultra 192G (Qwen3 32B Instruct) →
Common questions
What is the best local LLM for a Mac M3 Max 128G?
For everyday use, Llama 3.2 90B Vision Instruct at Q4_K_M — about 57.5 GB of the card's 128 GB at 4K context, leaving 38.5 GB for a longer window. If you want the largest thing that will load, that is DBRX Instruct at Q4_K_M (84.3 GB, 11.7 GB spare). "Best" here means best fit for the memory budget — this index does not run task benchmarks, so it cannot tell you which model is smarter.
Can a Mac M3 Max 128G run Qwen3 235B-A22B Instruct?
No. Its smallest build here, Q3_K_M, needs about 119.6 GB and this card has 96.0 GB usable — short by 23.6 GB before any context beyond 4K. The largest model this card does clear is DBRX Instruct.
How many tokens per second does a Mac M3 Max 128G do on an 8B model at Q4?
This index has no measured run on this card, so it does not publish a figure. What can be stated from specifications: generating a token requires reading every weight once, Llama 3.1 8B Instruct at Q4_K_M is 4.6 GB of weights, and this card moves 400 GB/s — a ceiling near 87 tok/s. Batch size, context length, the runtime and how much of the model sits in cache all take you below it.
Mac M3 Max 128G or Mac M5 Max 128G for local LLMs?
They hold the same models: 128 GB against 128 GB fits 100 and 100 of 110 respectively at 4K. The difference is throughput — 400 GB/s against 614 GB/s, a 1.5× gap in how fast the weights can be read, which is what token generation is bound by. The Mac M5 Max 128G generates faster on any model both can hold.
100 of 110 indexed models fit comfortably in 128GB at 4K context, each at the highest-quality quant that still leaves headroom.
This page's figures change when the model or the runtime does.Last updated 2026-10-18 RSS → /feed.xml