Apple · A18 Pro · Laptop
Best AI Models for MacBook Neo (A18 Pro, 8 GB)
Runs only small models; 9B doesn't fit in 4.5 GB.
| # | Model | Speed | llmrun Score | Quant · Memory |
|---|---|---|---|---|
| 1 | Qwen3.5 4BAlibabaHow to run → Q4_K_M · 3.4 GB · 12.9 t/s | 12.9 tok/s | Q4_K_M · 3.4 GB | |
| 2 | Qwen3.5 2BAlibaba Q4_K_M · 1.9 GB · 23.7 t/s | 23.7 tok/s | Q4_K_M · 1.9 GB | |
| 3 | MiniCPM5 2BOpenBMB Q4_K_M · 2.5 GB · 22.1 t/s | 22.1 tok/s | Q4_K_M · 2.5 GB | |
| 4 | LFM3 1.7BLiquid AI Q4_K_M · 2.1 GB · 29.2 t/s | 29.2 tok/s | Q4_K_M · 2.1 GB | |
| 5 | MiniCPM V 4.7 1BOpenBMB Q4_K_M · 1.6 GB · 32.8 t/s | 32.8 tok/s | Q4_K_M · 1.6 GB |
The MacBook Neo (A18 Pro, 8 GB) is a 8 GB laptop with more memory than speed: it runs only small models. macOS keeps about 3.5 GB, leaving 4.5 GB for models. Prompt reading is its weak spot: a 4K-token prompt waits 53 s before the first word on Qwen3.5 4B. At its $699 MSRP it gives 15.7 tok/s per $1,000, 5th of the 6 machines compared here. A18 Pro, released 2026.
MacBook Neo (A18 Pro, 8 GB) strengths and weaknesses
Weaknesses
- 7–9B models don't fit: Qwen3.5 9B needs 6.6 GB at Q4 vs 4.5 GB usable here
- Slow prompt reading: A 32K-token prompt on Qwen3.5 4B waits 7.1 min for the first token
- No CUDA: vLLM, AWQ, GPTQ and FP8 builds don't run; use GGUF or MLX
Memory for models
4.5 GB of 8 GB9B models need 6.6 GB at Q4, more than the 4.5 GB available.Or rent an RTX 3080 Ti (12 GB) from $0.13/hr →
Speed per dollar and per watt
On Qwen3.5 9B (9B params) at Q4_K_M. New means a shop price, either a street quote or the maker's own price; used is compared with used. The last two columns are the speed you get per $1,000 of price and per 100 W of power; higher is better. The MacBook Neo (A18 Pro, 8 GB) has no current new price, so it is shown at its MSRP.
| Machine | New price | tok/s | tok/s per $1,000 | tok/s per 100 W |
|---|---|---|---|---|
RTX 509032 GB | about $5,502 | 183 | 33.3 | 31.8 |
Mac Mini M616 GB | about $899 | 16.9 | 18.8 | – |
Mac Mini M4 Pro24 GB | about $1,599 | 30.0 | 18.8 | 21.4 |
Mac Mini M416 GB | about $799 | 13.2 | 16.5 | 20.3 |
| about $699 | 11.0 | 15.7 | – | |
MacBook Pro 14 M532 GB | about $1,839 | 16.9 | 9.2 | – |
Cost per token: the MacBook Neo (A18 Pro, 8 GB) against 8–32 GB machines and GPUs
On Qwen3.5 9B (9B params, Q4_K_M, 4K context). Speed, cost per token and power per token depend on the model.
- Machine, MSRP
- $4.03
- Power
- not published
Renting costs as much as buying after ≈8,500 hours (≈5.8 years at 4 h a day).
Each dot is a machine or GPU; hover or tap it for its numbers, click to open it. Machines without a price for the chosen basis are left out. Machines without a published power figure are left out of Per watt and costed without power. How we calculate cost per token
The same chip from other makers first, then machines and GPUs people weigh against it, each with the one reason it is worth a look. Speeds on Qwen3.5 9B.
Output speed depends on memory bandwidth (60 GB/s). Grey rows need more than the 4.5 GB available.
- 9BQwen3.5 9B · not at 4-bitneeds 6.6 GB
- 14BQwen3 14B · not at 4-bitneeds 11.9 GB
- 27–31BQwen3.8 27B · not at 4-bitneeds 17.9 GB
- 35B MoEQwen3.6 35B A3B · not at 4-bitneeds 22.0 GB
- 70BLlama 3.3 70B Instruct · not at 4-bitneeds 51.3 GB
- 120B MoEGPT OSS 120B · not at 4-bitneeds 70.8 GB
Estimated speeds: Q4_K_M, 4K context, one user. How we estimate
Unified-memory platforms: writing vs compute
- A18 Pro8 GB · 60 GB/s11.04
- M18 GB · 68.3 GB/s12.55
- M28 GB · 100 GB/s18.37
- Orin Nano8 GB · 102 GB/s10.417
- M38 GB · 102.4 GB/s18.76
Machines with about the same memory bandwidth write at about the same speed; compute decides how fast prompts are read. Speeds on Qwen3.5 9B.
All 604 models on the MacBook Neo (A18 Pro, 8 GB)
Featured onlyBest quant that fits in 4.5 GB, ordered by llmrun Score. Fit grades how well it runs here, S best. How ranking works
| Model | Quant | VRAM | Speed | Context | Score | Fit |
|---|---|---|---|---|---|---|
Qwen2.5 Coder 3B InstructChatCodeFunctions | Q4_K_M | 2.8 GB | 18.8 t/s | 33K | B58, Decent | |
| BF16 | 3 GB | 14.0 t/s | 2K | B53, Decent | ||
| Q4_K_M | 0.8 GB | 54.5 t/s | 131K | A81, Runs well | ||
Jan V3.5 4BChatMath | Q4_K_M | 4.5 GB | 13.4 t/s | 262K | D23, Barely runs | |
| Q3_K_L | 4.4 GB | 18.6 t/s | — | C32, Usable | ||
Pythia 1BChat | Q4_K_M | 0.7 GB | 59.2 t/s | 2K | A83, Runs well | |
| BF16 | 4.3 GB | 10.1 t/s | 262K | D25, Barely runs | ||
| Q4_K_M | 1.7 GB | 32.8 t/s | 33K | A72, Runs well | ||
Granite 3.0 1B A400m InstructChatFunctions | Q4_K_M | 1.3 GB | 71.5 t/s | 4K | S86, Runs great | |
Granite 4.0 MicroChatFunctions | Q4_K_M | 3.7 GB | 16.7 t/s | 131K | B54, Decent | |
LFM2 350M ExtractChatFunctions | Q4_K_M | 0.8 GB | 72.4 t/s | 128K | S86, Runs great | |
| Q4_K_M | 1.0 GB | 41.6 t/s | 2K | A76, Runs well | ||
| Q4_K_M | 1.2 GB | 60.9 t/s | 16K | A84, Runs well | ||
Phi 4 Mini ReasoningChatMathCodeReasoningFunctions | Q3_K_L | 4.4 GB | 16.6 t/s | 131K | D28, Barely runs | |
Codegen 350M MonoChatCode | FP16 | 0.8 GB | 54.5 t/s | 2K | A81, Runs well |
Specs
- Chip
- A18 Pro
- CPU cores
- 6
- GPU cores
- 5
- Unified memory
- 8 GB
- For models
- 4.5 GB (macOS keeps 3.5 GB)
- Memory bandwidth
- 60 GB/s
- FP16 (dense)
- 3.6 TFLOPS
- Neural Engine
- 35 TOPS
- NPU
- 16-core Neural Engine
- Size
- MacBook Neo
- Released
- 2026-03-11
- MSRP
- $699
MacBook Neo (A18 Pro, 8 GB) FAQ
Is the MacBook Neo (A18 Pro, 8 GB) worth it for local LLMs in 2026?
At its $699 MSRP it gives 15.7 tok/s per $1,000 on Qwen3.5 9B, 5th of the 6 machines compared here. It runs only small models; 9B doesn't fit in 4.5 GB.
How much memory is available for AI on the MacBook Neo (A18 Pro, 8 GB)?
About 4.5 GB of the 8 GB. macOS keeps roughly 3.5 GB for itself and other apps.
What is the largest model the MacBook Neo (A18 Pro, 8 GB) can run?
Qwen3.5 4B (4.7B) is the largest at 4-bit or better: at Q4_K_M it uses 3.4 GB, 76% of the available memory, and runs at about 12.9 tok/s.
Can the MacBook Neo (A18 Pro, 8 GB) run 9B models?
No. Qwen3.5 9B needs 6.6 GB at Q4, more than the 4.5 GB available.
How long does the MacBook Neo (A18 Pro, 8 GB) take to read a long prompt?
On Qwen3.5 4B, about 53 s for a 4K-token prompt and 7.1 min for a 32K-token document before the first word appears.
Is renting a GPU cheaper than buying the MacBook Neo (A18 Pro, 8 GB)?
An RTX 3080 Ti (12 GB) rents from $0.08/hr. Renting costs as much as buying after ≈8,500 hours (≈5.8 years at 4 h a day).