The verdict is the new Mac mini M5 Pro (48 GB) at $2,299, unless a used M4 Pro (48 GB) costs under about $2,000. The LLMCheck index puts Recursive Flint at 92 tok/s (community run) on it against 80 tok/s (community run) on the used Mac, and both fit the same 71 of the 130 models.
| M4 Pro (48 GB) | M5 Pro (48 GB) | |
|---|---|---|
| Chip | M4 Pro | M5 Pro |
| GPU cores | 16- or 20-core | 16- or 20-core |
| Memory bandwidth | 273 GB/s | 307 GB/s |
| Unified memory | 48 GB | 48 GB |
| Price | — | $2,299 |
| Price per GB | — | $48 |
| Models that fit (of 130) | 71 | 71 |
| Best local LLM here | KAT-Coder-V2.5 estimated66 tok/s | LFM2.5-24B-A2B community114 tok/s |
| Recursive Flint | community80 tok/s | community92 tok/s |
| Price per tok/s on Recursive Flint | — | $25 |
| Where to buy | Used / refurbished (Apple no longer sells it new) | New · build-to-order from Apple |
Every speed below is the LLMCheck index’s figure for that exact configuration — the same number its Mac mini M4 Pro (48 GB) and Mac mini M5 Pro (48 GB) pages print, with its provenance mark. Across the 69 shared models with a figure on both, the M5 Pro column is a median ×1.11 of the M4 Pro column.
| Model | Size | M4 Pro tok/s | M5 Pro tok/s | Ratio |
|---|---|---|---|---|
| KAT-Coder-V2.5 | 35B MoE | estimated66 | estimated72 | ×1.09 |
| Qwen3.8-27B | 27.8B | estimated13 slow | sourced17 | ×1.31 |
| Nanbeige4.3-33B | 33.1B | estimated11 slow | estimated12 slow | ×1.09 |
| Qwen 3.6-35B-A3B | 35B MoE | estimated74 | community79 | ×1.07 |
| Recursive Flint | 39.8B MoE | community80 | community92 | ×1.15 |
| LFM2.5-24B-A2B | 24B MoE | estimated90 | community114 | ×1.27 |
| Qwen 3.6-27B | 27B | estimated14 slow | estimated15 | ×1.07 |
| Nemotron 3.5 Lightning | 30B MoE | estimated67 | community72 | ×1.07 |
| MiniCPM-V 4.7 35B-A3B | 35.2B MoE | community75 | community57 | ×0.76 |
| North Mini Code 1.0 | 30.5B MoE | estimated65 | community70 | ×1.08 |
| Gemma 4 26B-A4B | 26B MoE | estimated72 | estimated78 | ×1.08 |
| Nemotron 3.5 Bolt 74B-A9B caveat | 74B MoE | estimated28 | estimated31 | ×1.11 |
| Muse Glimmer 30B | 30B | community10 slow | estimated11 slow | ×1.10 |
| GLM-4.7-Flash | 31B MoE | estimated68 | estimated73 | ×1.07 |
| Ripple-65B Preview | 64.8B | estimated6 slow | estimated7 slow | ×1.17 |
Nothing: both fit the same 71 of the 130 models in the LLMCheck index, because both have 48 GB. The difference is speed and price.
The new M5 Pro (48 GB) costs $2,299 (Apple list, 14 Nov 2026): $48 per GB and $25 per tok/s on Recursive Flint. A used M4 Pro (48 GB) matches that $25 per tok/s at about $2,000; 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 and the remaining years of macOS support.
Bottom line: The new M5 Pro (48 GB), unless the used one costs under about $2,000.
Renewed stock of the 48 GB Mac mini M4 Pro varies: check the memory in the listing before you buy. Amazon does not stock the Mac mini M5 Pro (48 GB): Apple builds it to order.
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 Mac mini M5 Pro (48 GB) at $2,299, unless a used M4 Pro (48 GB) costs under about $2,000. The LLMCheck index puts Recursive Flint at 92 tok/s (community run) on it against 80 tok/s (community run) on the used Mac, and both fit the same 71 of the 130 models.
Decode speed on Apple Silicon follows memory bandwidth: the Mac mini M5 Pro (48 GB) reads memory at 307 GB/s, the Mac mini M4 Pro (48 GB) at 273 GB/s. The LLMCheck index puts Recursive Flint at 92 tok/s (community run) against 80 tok/s (community run). Across the 69 shared models with a figure on both, the median gain is 1.11x.
Yes. Both fit the same 71 of the 130 models in the LLMCheck index.
About $2,000 or less. At that price it matches the new Mac's $25 per tok/s on Recursive Flint ($2,299 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 Mac mini M4 Pro (48 GB): KAT-Coder-V2.5 at ~66 tok/s (estimated). On the Mac mini M5 Pro (48 GB): LFM2.5-24B-A2B at 114 tok/s (community run). 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.