The best local LLM for a MacBook Air M1 (8 GB) is MiniCPM5-2B at ~39 tok/s (estimated). Its 8 GB fits 27 of the 130 models in the LLMCheck index, up to 9.41B-class. For everyday chat and coding, MiniCPM5-2B is the sweet spot.
Considering this Mac?
MacBook Air M5 (16GB) on Amazon → · compare every Mac for local AI →
Ranked by LLMCheck suitability (capability balanced against speed on the M1); the top pick is the best-ranked model that reaches 15 tok/s here, has a standard Mac runtime and fits the GPU memory limit (raised if need be), and slower models are tagged slow. Click a model for its full figures and setup. Every speed carries its provenance mark: most are index estimates from the memory-bandwidth model, a filled square is a published figure and a hollow circle a community run — submit a real run →
| # | Model | Size | License | Speed | 64K context | Capability |
|---|---|---|---|---|---|---|
| 1 | MiniCPM5-2B | 2.52B | Apache 2.0 | estimated39 tok/s | 21/50 | |
| 2 | Bonsai 27B over GPU limit | 27B | Apache 2.0 | — | 33/50 | |
| 3 | Ling-3.0-tiny over GPU limit | 7.9B MoE | MIT | estimated87 tok/s | 15/50 | |
| 4 | K2-Horizon-3.7B | 5.1B | Apache 2.0 | estimated18 tok/s | 22/50 | |
| 5 | LFM3-1.7B | 1.73B | LFM Open | estimated56 tok/s | agent-ready | 17/50 |
| 6 | Nanbeige4.2-3B | 3B | Apache 2.0 | estimated30 tok/s | 20/50 | |
| 7 | MiMo-V2.6-Distill-Qwen-9B slow over GPU limit | 9.41B | MIT | estimated10 tok/s | 21/50 | |
| 8 | Spark-X2.6-4B | 4.11B | Apache 2.0 | estimated18 tok/s | 20/50 | |
| 9 | Maple Preview 20B-A1B over GPU limit | 20B MoE | MIT | — | 20/50 | |
| 10 | Nemotron 3.5 Bolt E4B over GPU limit | 4B | OpenMDW | estimated19 tok/s | 17/50 | |
| 11 | Qwen 3.5 9B slow over GPU limit | 9B | Apache 2.0 | estimated10 tok/s | 18/50 | |
| 12 | Gemma 4 E4B over GPU limit | 4B | Apache 2.0 | estimated23 tok/s | 16/50 |
Showing the top 12 of 27 models that fit in 8 GB. See the full leaderboard or all benchmarks.
Context: “fits” means the weights at a short prompt. Coding agents (Claude Code, OpenCode, Codex CLI) hold 64K tokens or more, and every token adds KV cache: agent-ready marks a model whose weights plus a 64K-token 16-bit KV cache still fit in the 5.3 GB this Mac’s GPU can use with the limit raised (4 of the 24 models here with a KV figure). That is an estimate from each model’s config.json, not a measurement; a q8 KV cache halves the cache part, and — means the index has no KV figure for the model yet. The GPU memory limit below applies on top.
GPU memory limit: macOS lets the GPU use about 5.3 GB of this Mac’s 8 GB by default on macOS 15 and macOS 26 alike. Of the 27 models that fit, 11 need more than that at a short prompt and are tagged over GPU limit. An 8 GB Mac has no headroom to raise it safely: macOS needs the rest. How the limit is read →
The index has no verified Ollama tag for MiniCPM5-2B, so start from its listing and copy the exact name there: Hugging Face repo → · model card →
GGUF builds run in LM Studio or llama.cpp; MLX builds run in LM Studio or mlx-lm.
Prefer a GUI? LM Studio gives you a one-click download and chat window. For step-by-step help see our Ollama install guide, or open the MiniCPM5-2B on M1 page for its figures and setup.
Twice as fast: the cheapest Mac in the index that runs MiniCPM5-2B at 2× or more of this Mac’s estimated~39 tok/s is the Mac mini M2 Pro (16 GB) at estimated~104 tok/s, about $700 refurbished or used: search Amazon Renewed listings →. New: the Mac mini M6 (16 GB) at estimated~82 tok/s, $899 new (Apple list price, checked 2026-11-14): Mac mini M6 (16GB) on Amazon →.
For less: no cheaper Mac in the index runs MiniCPM5-2B at 15+ tok/s.
Speeds are each Mac’s own figure from the index, with its provenance mark. Prices come from data/macs.json: Apple’s list price for a new Mac (dated), an approximate street price for a used or refurbished one. Amazon stock checked 2026-11-18. Amazon links are affiliate links: as an Amazon Associate, LLMCheck earns from qualifying purchases.
The MacBook Air M1 (8 GB) tops out at MiMo-V2.6-Distill-Qwen-9B. A current MacBook Air M5 (16 GB) fits 11 more of the index’s models and reads memory at 153 GB/s instead of 68 GB/s, so models also generate faster:
As an Amazon Associate, LLMCheck earns from qualifying purchases. Affiliate links cost you nothing extra and never influence our rankings.
LLMCheck. (2026). Best Local LLMs for the MacBook Air M1 (8 GB). llmcheck.net. https://llmcheck.net/best-llm/macbook-air-m1-8gb/ (updated ). CC BY 4.0.
MiniCPM5-2B (2.52B, Apache 2.0) is the best all-round pick at ~39 tok/s (estimated) on the MacBook Air M1. If you want maximum speed, Ling-3.0-tiny hits ~87 tok/s (estimated); for maximum capability, Bonsai 27B still fits in 8 GB.
About 27 of the 130 models in the LLMCheck index fit in 8 GB of unified memory, from compact models up to MiMo-V2.6-Distill-Qwen-9B (9.41B).
No. A dense 70B such as Llama 3.3 70B needs about 44 GB at 4-bit; with 8 GB, stick to models like MiniCPM5-2B. For a 70B, look at a 64 GB+ Mac.
8 GB is enough for dense models up to about 9B at 4-bit (the largest in the index that fits is MiMo-V2.6-Distill-Qwen-9B), and for mixture-of-experts models such as Ling-3.0-tiny (~87 tok/s (estimated)). As a rule of thumb a 4-bit model needs about 0.57 GB per billion parameters plus ~2 GB, and macOS needs its own share of memory too. Because Apple Silicon uses unified memory, the same pool serves as system RAM and GPU memory.