The verdict is the new MacBook Air M5 (24 GB) at $1,499 (Apple list, 14 Nov 2026), unless a used M4 (24 GB) costs under about $1,250. The LLMCheck index puts Qwen 3.6-35B-A3B at ~59 tok/s (estimated) on the new Mac against ~50 tok/s (estimated) on the used one, and both fit the same 63 of the 130 models.
| M4 (24 GB) | M5 (24 GB) | |
|---|---|---|
| Chip | M4 | M5 |
| GPU cores | 10-core | 10-core |
| Memory bandwidth | 120 GB/s | 153 GB/s |
| Unified memory | 24 GB | 24 GB |
| Price | — | $1,499 |
| Price per GB | — | $62 |
| Models that fit (of 130) | 63 | 63 |
| Best local LLM here | Qwen 3.6-35B-A3B estimated50 tok/s | Qwen 3.6-35B-A3B estimated59 tok/s |
| Qwen 3.6-35B-A3B | estimated50 tok/s | estimated59 tok/s |
| Price per tok/s on Qwen 3.6-35B-A3B | — | $25 |
| Where to buy | Used / refurbished (Apple no longer sells it new) | New · sold new by Apple (Amazon stock not verified) |
Every speed below is the LLMCheck index’s figure for that exact configuration — the same number its MacBook Air M4 (24 GB) and MacBook Air M5 (24 GB) pages print, with its provenance mark. Across the 59 shared models with a figure on both, the M5 column is a median ×1.25 of the M4 column.
| Model | Size | M4 tok/s | M5 tok/s | Ratio |
|---|---|---|---|---|
| Qwen 3.6-35B-A3B | 35B MoE | estimated50 | estimated59 | ×1.18 |
| Qwen3.8-27B | 27.8B | estimated7 slow | estimated8 slow | ×1.14 |
| Nanbeige4.3-33B | 33.1B | estimated5 slow | estimated6 slow | ×1.20 |
| Qwen 3.6-27B | 27B | estimated6 slow | estimated8 slow | ×1.33 |
| Recursive Flint | 39.8B MoE | estimated56 | estimated67 | ×1.20 |
| KAT-Coder-V2.5 | 35B MoE | estimated36 | estimated44 | ×1.22 |
| Muse Glimmer 30B | 30B | estimated4 slow | estimated6 slow | ×1.50 |
| LFM2.5-24B-A2B | 24B MoE | community59 | community74 | ×1.25 |
| Gemma 4 31B | 31B | estimated6 slow | estimated7 slow | ×1.17 |
| MiniCPM-V 4.7 35B-A3B | 35.2B MoE | estimated44 | estimated52 | ×1.18 |
| Nemotron 3.5 Lightning | 30B MoE | estimated34 | estimated41 | ×1.21 |
| Gemma 4 26B-A4B | 26B MoE | estimated40 | estimated48 | ×1.20 |
| GLM-4.7-Flash | 31B MoE | estimated37 | estimated45 | ×1.22 |
| Raon-1 mini preview | 28.8B MoE | estimated45 | estimated54 | ×1.20 |
| North Mini Code 1.0 | 30.5B MoE | estimated28 | community34 | ×1.21 |
Nothing: both fit the same 63 of the 130 models in the LLMCheck index, because both have 24 GB. The difference is speed and price.
The new M5 (24 GB) costs $1,499 (Apple list, 14 Nov 2026): $62 per GB and $25 per tok/s on Qwen 3.6-35B-A3B. A used M4 (24 GB) matches that $25 per tok/s at about $1,250; below that price it is the better value on decode speed alone. The index records no used prices: check current listings, and weigh the warranty, battery health and the remaining years of macOS support.
Bottom line: The new M5 (24 GB), unless the used one costs under about $1,250.
Renewed stock of the 24 GB MacBook Air M4 varies: check the memory in the listing before you buy. The MacBook Air M5 (24 GB) is sold new by Apple; Amazon stock of the 24 GB configuration is not verified, so check the listing’s memory.
As an Amazon Associate, LLMCheck earns from qualifying purchases. Affiliate links cost you nothing extra and never influence the verdicts.
Each speed is decode (generation) tok/s from the LLMCheck index for that exact configuration: the lower-median measured run on that chip where one exists (a filled square is a published figure, a hollow circle a community run), otherwise an estimate from the memory-bandwidth model (hollow diamond, printed with a tilde). LLMCheck runs no tests of its own. Prompt processing (prefill) depends on GPU compute rather than bandwidth, and the index does not estimate it; Apple’s M5-generation GPUs add per-core Neural Accelerators aimed at that work.
The verdict is the new MacBook Air M5 (24 GB) at $1,499 (Apple list, 14 Nov 2026), unless a used M4 (24 GB) costs under about $1,250. The LLMCheck index puts Qwen 3.6-35B-A3B at ~59 tok/s (estimated) on the new Mac against ~50 tok/s (estimated) on the used one, and both fit the same 63 of the 130 models.
Decode speed on Apple Silicon follows memory bandwidth: the MacBook Air M5 (24 GB) reads memory at 153 GB/s, the MacBook Air M4 (24 GB) at 120 GB/s. The LLMCheck index puts Qwen 3.6-35B-A3B at ~59 tok/s (estimated) against ~50 tok/s (estimated). Across the 59 shared models with a figure on both, the median gain is 1.25x.
Yes. Both fit the same 63 of the 130 models in the LLMCheck index.
About $1,250 or less. At that price it matches the new Mac's $25 per tok/s on Qwen 3.6-35B-A3B ($1,499 at Apple list, checked 14 Nov 2026). The LLMCheck index records no used prices, so check current listings, and weigh warranty and macOS support.
On the MacBook Air M4 (24 GB): Qwen 3.6-35B-A3B at ~50 tok/s (estimated). On the MacBook Air M5 (24 GB): Qwen 3.6-35B-A3B at ~59 tok/s (estimated). These are the headline picks of the LLMCheck Best by Mac pages: the best-ranked model that reaches 15 tok/s on that Mac and has a standard Mac runtime.