The verdict is the Mac mini M5 Pro (24 GB): the LLMCheck index gives both Macs the same speed on 62 of the 63 shared models, and it runs the same headline pick, LFM2.5-24B-A2B, at 114 tok/s (community run). The $600 extra for the M5 Pro (48 GB) buys 8 more models, led by Qwen3-Coder-Next.
| M5 Pro (24 GB) | M5 Pro (48 GB) | |
|---|---|---|
| Chip | M5 Pro | M5 Pro |
| GPU cores | 16- or 20-core | 16- or 20-core |
| Memory bandwidth | 307 GB/s | 307 GB/s |
| Unified memory | 24 GB | 48 GB |
| Price | $1,699 | $2,299 |
| Price per GB | $71 | $48 |
| Models that fit (of 130) | 63 | 71 |
| Best local LLM here | LFM2.5-24B-A2B community114 tok/s | LFM2.5-24B-A2B community114 tok/s |
| LFM2.5-24B-A2B | community114 tok/s | community114 tok/s |
| Price per tok/s on LFM2.5-24B-A2B | $15 | $20 |
| Where to buy | New · Amazon stocks this configuration | 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 M5 Pro (24 GB) and Mac mini M5 Pro (48 GB) pages print, with its provenance mark. Across the 63 shared models with a figure on both, the 48 GB column is a median ×1.00 of the 24 GB column.
| Model | Size | 24 GB tok/s | 48 GB tok/s | Ratio |
|---|---|---|---|---|
| LFM2.5-24B-A2B | 24B MoE | community114 | community114 | ×1.00 |
| Qwen3.8-27B | 27.8B | sourced17 | sourced17 | ×1.00 |
| Recursive Flint | 39.8B MoE | estimated92 | community92 | ×1.00 |
| KAT-Coder-V2.5 | 35B MoE | estimated72 | estimated72 | ×1.00 |
| Qwen 3.6-35B-A3B | 35B MoE | estimated79 | community79 | ×1.00 |
| Nanbeige4.3-33B | 33.1B | estimated12 slow | estimated12 slow | ×1.00 |
| Qwen 3.6-27B | 27B | estimated15 | estimated15 | ×1.00 |
| Nemotron 3.5 Lightning | 30B MoE | community72 | community72 | ×1.00 |
| North Mini Code 1.0 | 30.5B MoE | community70 | community70 | ×1.00 |
| Gemma 4 26B-A4B | 26B MoE | estimated78 | estimated78 | ×1.00 |
| Raon-1 mini preview | 28.8B MoE | community86 | community86 | ×1.00 |
| GLM-4.7-Flash | 31B MoE | estimated73 | estimated73 | ×1.00 |
| Muse Glimmer 30B | 30B | estimated11 slow | estimated11 slow | ×1.00 |
| Laguna XS 2.1 | 33B MoE | estimated83 | estimated83 | ×1.00 |
| Gemma 4 31B | 31B | estimated13 slow | estimated13 slow | ×1.00 |
A model fits when its minimum RAM (its smallest build of at least 3 bits with a working Mac runtime, plus about 2 GB for the runtime and a short context; 2-bit only where a published Mac run used it) is at most the Mac’s unified memory. The extra memory unlocks these models from the LLMCheck index:
| Model | Size | Min RAM | 48 GB tok/s |
|---|---|---|---|
| Nemotron 3.5 Bolt 74B-A9B caveat | 74B MoE | 45 GB | estimated31 |
| Ripple-65B Preview | 64.8B | 33 GB | estimated7 |
| Qwen3-Coder-Next | 80B MoE | 48 GB | estimated49 |
| DeepSeek R1 70B | 70B | 44 GB | estimated6 |
| Apertus 1.5 70B | 70B | 44 GB | estimated6 |
| Hermes 4 70B | 70B | 38 GB | estimated6 |
| Qwen 2.5 72B | 72B | 46 GB | estimated6 |
| Llama 3.3 70B | 70B | 44 GB | estimated6 |
At Apple’s list prices the M5 Pro (24 GB) costs $71 per GB of unified memory and the M5 Pro (48 GB) $48. On LFM2.5-24B-A2B — the reference model here, which both run — that is $15 and $20 per tok/s of decode speed: the M5 Pro (24 GB) buys more speed per dollar.
The $600 premium (35% more) buys no extra decode speed on LFM2.5-24B-A2B and 8 more models that fit.
Bottom line: The M5 Pro (24 GB): same speeds; the extra memory buys 8 more models.
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.
The verdict is the Mac mini M5 Pro (24 GB): the LLMCheck index gives both Macs the same speed on 62 of the 63 shared models, and it runs the same headline pick, LFM2.5-24B-A2B, at 114 tok/s (community run). The $600 extra for the M5 Pro (48 GB) buys 8 more models, led by Qwen3-Coder-Next.
No. Both use the M5 Pro at 307 GB/s, and decode speed on Apple Silicon follows memory bandwidth, so the LLMCheck index gives the same figure on 62 of the 63 shared models (LFM2.5-24B-A2B: 114 tok/s (community run)). The extra 24 GB buys models and context, not speed.
8 models in the LLMCheck index need more than 24 GB but fit in 48 GB, led by Nemotron 3.5 Bolt 74B-A9B, Ripple-65B Preview, Qwen3-Coder-Next. Both run the other 63.
It costs $2,299 against $1,699 (Apple list prices, checked 14 Nov 2026). For that it gives the same 114 tok/s (community run) on LFM2.5-24B-A2B and fits 8 more models. For local LLMs the cheaper Mac is the better buy; pay more for the extra 24 GB of headroom, not for speed.
On the Mac mini M5 Pro (24 GB): LFM2.5-24B-A2B at 114 tok/s (community run). 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.