Apple M6 (16 GB, 153 GB/s) for local LLMs
Written by Jakub Rusinowski · Last updated
With 16 GB of LPDDR5X at 153 GB/s, the M6 (16 GB, 153 GB/s) runs 88 catalogued models at Q4_K_M with 8K context. The largest that fits is Phi-4 Reasoning Vision 15B (~11.6 GB), and the top pick is Mellum2.1 12B Thinking at 29–60 tok/s.
The 16 GB configuration of the M6. Apple gives it 153 GB/s of memory bandwidth; the 24 GB and 32 GB configurations get 170 GB/s. Same 12-core CPU and 12-core GPU, so the difference is memory speed, which sets decode speed.
Models that run on the M6 (16 GB, 153 GB/s)
Q4_K_M, 8K context, 12 GB usable. Ranked by quality and speed.
| Model | Memory · marker = 12 GB | VRAM | Speed |
|---|---|---|---|
| Mellum2.1 12B Thinking Mellum2.1 | 9 GB | 29–60 tok/s | |
| LFM2.5-8B-A1B LFM2.5 | 5.8 GB | 23–49 tok/s | |
| Gemma 3n E4B Gemma 3n | 5.5 GB | 15–30 tok/s | |
| Gemma 3n E2B Gemma 3n | 4.1 GB | 23–47 tok/s | |
| Gemma 4 E2B Gemma 4 | 3.9 GB | 13–26 tok/s | |
| K2 Horizon 3.7B K2 Horizon | 5.1 GB | 12–25 tok/s | |
| Bielik 4.5B v3.0 Instruct Bielik | 4.2 GB | 15–31 tok/s | |
| Ternary Bonsai 2 4B Ternary Bonsai 2 | 3.6 GB | 18–37 tok/s | |
| StarCoder 2 7B StarCoder 2 | 5.7 GB | 11–22 tok/s | |
| Cosmos 3 Nano Cosmos 3 | 10.5 GB | 8.4–17 tok/s | |
| Qwen 3.5 9B Qwen 3.5 | 6.6 GB | 9.1–19 tok/s | |
| Olmo 3 7B Olmo 3 | 5 GB | 9.8–20 tok/s |
Buy it or rent the same memory
Buy the card, or rent a GPU with the same memory by the hour to try models first.
or compare on Vast.ai from $0.35/hr (typical low · varies)
As an Amazon Associate we earn from qualifying purchases. Cloud GPU links are referral links — we may earn a commission at no extra cost to you.
Speed vs other GPUs
Llama 3.1 8B, Q4_K_M. Estimated ranges. How this is calculated
Specifications
Specs last updated 2026-11-11.
- Memory
- 16 GB LPDDR5X
- Memory bandwidth
- 153 GB/s
- Architecture
- ARM, 2nm TSMC
- Series
- Apple Silicon
- Release year
- 2026
- Compute backends
- METAL
- Usable for models
- 12 GB (~75% of unified)
Similar GPUs
Frequently asked questions
Can the Apple M6 (16 GB, 153 GB/s) run local LLMs?
Yes. With 16 GB (12 GB usable by a model) it runs 88 of the catalogued models at Q4_K_M with 8K context; the largest is Phi-4 Reasoning Vision 15B, needing about 11.6 GB.
How fast is the Apple M6 (16 GB, 153 GB/s) for AI inference?
It is estimated to run Llama 3.1 8B at 9.7–20 tok/s at Q4_K_M. Llama 3.3 70B does not fit: it needs about 44 GB against 12 GB usable. These are modelled estimates from memory bandwidth, not measurements; the methodology page shows the formula.
What LLMs can I run on 16 GB?
Among the best that fit: Mellum2.1 12B Thinking, LFM2.5-8B-A1B, Gemma 3n E4B, Gemma 3n E2B, Gemma 4 E2B.