Can the M6 run a 70B model?

Not comfortably. A 70B model at 4-bit needs about 40 GB, and the M6 has at most 32 GB. Models up to about 42B parameters fit well.

How many tokens per second does the M6 get?

For an 8B model at 4-bit, the ceiling is about 38 tokens a second, and real runs reach roughly half to four fifths of it. Bigger models are proportionally slower.

Is the M6 good for AI?

For running models locally, memory and bandwidth matter most. The M6 has up to 32 GB at 170 GB/s, enough for models up to about 42 billion parameters. It can also earn by running AI work for MLX Network with MLX Node.

M6 for local LLMs: memory, speed and what fits

1The specs that matter for AI

Unified memory	32 GB
Memory bandwidth	170 GB/s, or 153 GB/s on the lower configuration
Found in	Mac mini (2026)
Dedicated seat for it	Core Latest seat: mines up to 30 MLX a day

2Which models fit

At 4 bits per weight a model needs a little over half a gigabyte per billion parameters, plus room for its context and macOS. With 32 GB, the M6 comfortably holds models up to about 42 billion parameters.

Model at 4-bit	Memory it needs	Fits on the M6
8B model	4.5 GB	Yes
14B model	7.9 GB	Yes
32B model	18 GB	Yes
70B model	39.4 GB	No
gpt-oss-120b, mixture-of-experts	65 GB	No
Qwen3.5-397B-A17B, mixture-of-experts	223.3 GB	No
DeepSeek-V4.1-Flash, mixture-of-experts	316 GB	No

3How fast it runs

Writing text is limited by memory bandwidth: every new token reads the model's active weights once. Dividing the M6's 170 GB/s by the bytes read per token gives the ceiling below. Real runs usually reach about 45% to 80% of it (llama.cpp benchmarks).

Model at 4-bit	Read per token	Ceiling for one chat
8B model	4.5 GB	38 tokens/s
14B model	7.9 GB	22 tokens/s
32B model	18 GB	9 tokens/s

M5-generation chips add neural accelerators to the GPU, so they read prompts several times faster than M4: Apple measured the first token arriving 3.3 to 4.1 times sooner with MLX (Apple).

5Compare with other chips

6Questions

Can the M6 run a 70B model?: Not comfortably. A 70B model at 4-bit needs about 40 GB, and the M6 has at most 32 GB. Models up to about 42B parameters fit well.
How many tokens per second does the M6 get?: For an 8B model at 4-bit, the ceiling is about 38 tokens a second, and real runs reach roughly half to four fifths of it. Bigger models are proportionally slower.
Is the M6 good for AI?: For running models locally, memory and bandwidth matter most. The M6 has up to 32 GB at 170 GB/s, enough for models up to about 42 billion parameters. It can also earn by running AI work for MLX Network with MLX Node.

Sources

Apple: M6 and M5 Ultraapple.com