Run on your Mac

Run Outlier Nano on MacBook Air M1

Last updated 2026-06-18 · Outlier v1.11.469

What does Outlier Nano run like on MacBook Air M1?

The pairing under examination is Outlier Nano (2.37 GB MLX 4-bit, Qwen3.5-4B base, 32K default context, 6 GB unified memory minimum) on a MacBook Air M1 (Apple M1, 8 CPU cores, 7-8 GPU cores, 8|16 GB unified memory, 68.25 GB/s, 2020). Outlier loads this tier as one mlx_lm process inside the bundled FastAPI sidecar. Apple Silicon decode on dense 4-bit weights is bandwidth-bound, so the ceiling on this exact MacBook Air M1 scales as the ratio 68.25 / 800 = 0.09× the published M1 Ultra number.

How much memory does the model take during generation?

Outlier’s Outlier Nano has a measured peak generation footprint of about 3 GB. The MacBook Air M1 base configuration ships with 8|16 GB unified, and the rule of thumb is to leave roughly 4 GB for the OS and one open browser tab. On MacBook Air M1 the Outlier Nano bandwidth-scaled estimate is ~6.1 tok/s [estimated from family-bandwidth ratio]; derivation: 71.7 tok/s on M1 Ultra × (68.25 / 800) bus ratio for the Apple M1. Treat as a first-order projection — the Outlier Nano number on this 2020 machine has not been formally measured.

What is the install path on MacBook Air M1 for Outlier Nano?

The 2020 MacBook Air M1 ships with 8|16 GB of unified memory, so headroom for the 6 GB Outlier Nano requirement is 2 GB on the base SKU. Step-by-step install instructions live on the install guide; the part that varies for this Apple M1 machine is that the 2.37 GB pull from Outlier-Ai/Outlier-Nano-4B-MLX-4bit over HTTPS lands in ~/Library/Application Support/Outlier/models/nano/ and download time is bandwidth-limited by the network, not by the 8-core Apple M1 CPU.

What context window can I use on a 8 GB Mac?

The Outlier Nano tier defaults to 32K context and caps at 256K. KV cache scales linearly with context length on dense models, so longer contexts trade headroom for capacity. On a 8|16 GB MacBook Air M1, the default context is the safe starting point.

What works well on this pairing, and what is still rough?

• The MacBook Air M1 (Apple M1, 68.25 GB/s) is bandwidth-bound on dense 4-bit decode; the throughput ceiling for Outlier Nano on this exact machine is approximately 68.25 divided by 2.37 GB tokens per second.
• The 16 GB base SKU of the MacBook Air M1 swaps aggressively under the Outlier Nano tier; close Chrome and other RAM-heavy applications before generation.
• The Apple M1 memory bus runs slower than M2 and later silicon in the same class; the Outlier Nano tier numbers are correspondingly lower than the M4 equivalent.
• Cold boot of Outlier on a 8-core Apple M1 is roughly 50 seconds while the bundled Python.framework warms; this is documented on the v1.9 backlog under the boot-time investigation thread.

What is the unique number for Outlier Nano on MacBook Air M1?

2.37 GB on disk against 8|16 GB unified memory means the weights alone consume about 30% of the base SKU’s RAM on the MacBook Air M1.

Should I pick a different tier?

For heavier work, the Lite tier is the next step up; check that the 8|16 GB SKU you have can fit its weights before installing. The unified-memory explainer works the bandwidth math out, and the bandwidth ratio for MacBook Air M1 is 0.09× M1 Ultra reference.

Download Outlier for Mac

Requires Apple Silicon (M1, M2, M3, or M4) — Intel Macs are not supported. macOS 12+.