The best MacBook Pro for local AI in October 2026 is the new 14-inch M6 with 32 GB at £2,399 if you want 27B models, or the 14-inch M5 Max with 128 GB at £6,699 if you want 70B models at full speed. Memory decides what runs. The new OLED model costs £400 more and adds no AI speed. Below: what each configuration runs, what it costs in pounds, how bandwidth sets tokens per second, and whether to buy now.
Key facts
- What is new: a 14-inch MacBook Pro with M6 (27 October 2026) and an OLED touchscreen MacBook Pro with M5 Pro or M5 Max (28 October). Both arrive from 6 November.
- UK prices (29 October 2026, including VAT): 14-inch M6 from £1,999 (16 GB, 1 TB); 14-inch OLED from £2,899; 128 GB from £6,699 (14-inch mini-LED) or £7,099 (14-inch OLED). The 16-inch with 128 GB costs £6,999 or £7,399.
- Memory options: M6 16, 24 or 32 GB; M5 Pro 24, 48 or 64 GB; M5 Max 36, 48, 64 or 128 GB. You cannot add memory later.
- Memory bandwidth (Apple): M6 153-170 GB/s; M5 Pro 307 GB/s; M5 Max 460 or 614 GB/s.
- Memory rule: 16 GB runs 8B models, 32 GB runs 27B, 64 GB runs 70B at 4-bit, and 128 GB runs 70B at 8-bit and models up to about 120B.
- Best value for local AI: 14-inch M6, 32 GB, 1 TB, £2,399.
- OLED or mini-LED: same chips, memory and bandwidth; OLED costs £400 more. M5 mini-LED prices are unchanged.
On this page
- Which MacBook Pro for local AI should you buy?
- What did Apple launch in October 2026?
- How much does the new MacBook Pro cost in the UK?
- How much memory does a MacBook Pro need for local AI?
- Does memory bandwidth change tokens per second?
- M6 vs M5 Pro vs M5 Max: which chip is best for AI?
- OLED or mini-LED: which MacBook Pro is better for local AI?
- Can you still buy the M5 MacBook Pro, and for how much?
- Should you buy the new MacBook Pro now or wait?
- Frequently asked questions
- Sources checked
Which MacBook Pro for local AI should you buy?
Buy the 14-inch MacBook Pro with M6 and 32 GB at £2,399 if 27B models are enough, and the 14-inch M5 Max with the 40-core GPU and 128 GB at £6,699 if you want 70B models at their best. In between, the 14-inch M5 Pro with 64 GB at £3,699 is the cheapest MacBook Pro that holds a 70B model.
Every figure here follows the rule on which Mac or mini PC to buy for local AI: buy memory first, speed second. A bigger chip cannot load a model that does not fit, and Apple solders the memory, so the choice at checkout is final.
- Pick the biggest model you want to run. Chat and drafts need 8B-14B models; coding help and agents work best with 27B-32B; the strongest home models are 70B and up.
- Buy the memory that model needs. The local LLM memory calculator does the sums for any model.
- Pick the cheapest chip that offers that memory. Up to 32 GB that is the M6. For 48 or 64 GB it is the M5 Pro. Only the 40-core M5 Max offers 128 GB.
- Pay for speed only if you will feel it. In the llama.cpp community benchmark, the 40-core M5 Max wrote more than three times faster than an M6 with an 8-core GPU, which matters most with big models.
- Choose the screen last. The OLED version costs £400 more on every M5 Pro and M5 Max configuration and runs AI at the same speed.
Will it stay on a desk? Then a laptop is the expensive route. The Mac mini M6 with 32 GB costs £1,299 (with 256 GB of storage) with the same M6 chip and 170 GB/s memory, £1,100 less than the 32 GB MacBook Pro with 1 TB.
What did Apple launch in October 2026?
Apple launched two MacBook Pro lines in two days. On 27 October it announced a 14-inch MacBook Pro with the new M6 chip, and on 28 October, at an event whose invitation carried the tagline "Pro and tell" (as reported by 9to5Mac; Apple's Newsroom does not use the name), it announced a MacBook Pro with PROLED XDR Display, a tandem OLED model that Apple says brings touch to the Mac for the first time.
- 14-inch MacBook Pro with M6. 12-core CPU, 12-core GPU with Neural Accelerators, a Dual 16-core Neural Engine, and 16 GB to 32 GB of memory at up to 170 GB/s. Mini-LED display, three Thunderbolt 4 ports.
- MacBook Pro with PROLED XDR Display. 14.2-inch or 16.3-inch tandem OLED screen with up to 2,000 nits peak brightness, touch and a Dynamic Island. M5 Pro or M5 Max, up to 128 GB of memory and up to 16 TB of storage.
- Dates. Pre-orders for both opened on 28 October 2026. Apple says deliveries and in-store stock begin on Friday 6 November.
- Unchanged: the M5 Pro and M5 Max chips date from March 2026. There is no M6 Pro or M6 Max yet.
Apple's own AI claim for the M6 is narrow. It says the M6 MacBook Pro is up to 1.3 times faster than the M5 model at time to first token, measured with a 14B model at 4-bit, a 16K-token prompt and LM Studio. That is how quickly a model starts answering a long prompt, not how fast it writes.
How much does the new MacBook Pro cost in the UK?
The new MacBook Pro costs from £1,999 in the UK for the 14-inch M6 with 16 GB and 1 TB, and £2,399 with 32 GB. The 14-inch OLED model starts at £2,899 and the 16-inch OLED at £3,399. The cheapest 128 GB MacBook Pro is still the mini-LED 14-inch M5 Max at £6,699; the OLED version is £7,099.
Prices from Apple's UK store on 29 October 2026, including VAT.
| 14-inch configuration | Mini-LED (Liquid Retina XDR) | OLED (PROLED XDR) | Runs well at 4-bit |
|---|---|---|---|
| M6, 16 GB, 1 TB | £1,999 | Not offered | 8B models; 14B at a squeeze |
| M6, 24 GB, 1 TB | £2,199 | Not offered | 14B models |
| M6, 32 GB, 1 TB | £2,399 | Not offered | 27B models such as Qwen3.8 27B |
| M5 Pro 15-core CPU, 24 GB, 1 TB | £2,499 | £2,899 | 14B models |
| M5 Pro 15-core CPU, 48 GB, 1 TB | £3,099 | £3,499 | 32B models, or 27B at 8-bit |
| M5 Pro 18-core CPU, 20-core GPU, 64 GB, 1 TB | £3,699 | £4,099 | 70B models |
| M5 Max 32-core GPU, 36 GB, 2 TB | £4,099 | £4,499 | 27B-32B models |
| M5 Max 40-core GPU, 64 GB, 2 TB | £5,099 | £5,499 | 70B models, faster |
| M5 Max 40-core GPU, 128 GB, 2 TB | £6,699 | £7,099 | 70B at 8-bit, models up to about 120B |
The 16-inch comes only with M5 Pro or M5 Max. It starts at £2,999 (M5 Pro, 24 GB) as a mini-LED model and £3,399 as OLED. With 64 GB it is £3,999 or £4,399, and the 128 GB M5 Max is £6,999 or £7,399.
Two costs to add. No MacBook Pro comes with a power adapter in the UK. Apple sells the 70 W for £59, the 96 W for £79 and the 140 W for £99; the 16-inch needs the 140 W to fast charge. Students pay £1,899 for the base M6 and £2,299 for the 32 GB M6 in Apple's UK education store. If you are buying before Christmas, the Black Friday AI deals page tracks UK price drops against these list prices.
How much memory does a MacBook Pro need for local AI?
A MacBook Pro needs 16 GB for 8B models, 32 GB for 27B models, 64 GB for 70B models at 4-bit and 128 GB for 70B at 8-bit or models up to about 120B. That is the same rule as the memory calculator and the rest of this site.
The sums are simple. A 4-bit model file is about 0.6 GB per billion parameters, so Llama 3.3 70B is about 42.4 GB and Qwen3.8 27B about 16.4 GB. A long conversation adds 2-4 GB on top. By default macOS lets the graphics side use about 75% of memory, so a 32 GB Mac gives a model roughly 21-24 GB and a 128 GB Mac about 96 GB.
Three spec-sheet details matter. The 16 GB M6 has slower memory (153 GB/s) than the 24 GB and 32 GB versions (170 GB/s). 64 GB on the M5 Pro needs the 20-core GPU. 128 GB needs the 40-core M5 Max.
Agents that read long documents need more headroom than a quick question. The setup for each memory size is in run a local LLM on your Mac.
Does memory bandwidth change tokens per second?
Yes. When a model writes, the chip reads the whole model from memory for every word, so writing speed rises with memory bandwidth and falls with model size. In the llama.cpp benchmark thread below, the 40-core M5 Max (614 GB/s) wrote more than three times faster than an M6 with an 8-core GPU (170 GB/s).
The cleanest comparison is the llama.cpp project's Apple silicon benchmark thread. It runs the same small model, Llama 2 7B at 4-bit, on every chip. These are community results, and nobody had published figures for the 12-core M6 MacBook Pro by 29 October 2026.
| Chip | Memory bandwidth (Apple) | Writing speed, Llama 2 7B Q4_0 (llama.cpp thread) | Prompt reading, same test |
|---|---|---|---|
| M6 | 153 GB/s (16 GB) or 170 GB/s | About 36 tokens a second (an M6 with an 8-core GPU and 170 GB/s, not the 12-core GPU in this MacBook Pro) | About 989 tokens a second (same 8-core M6) |
| M5 (old base 14-inch) | 153 GB/s | About 32 tokens a second | About 723 tokens a second |
| M5 Pro | 307 GB/s | About 66-68 tokens a second | About 1,340-1,621 tokens a second |
| M5 Max, 32-core GPU | 460 GB/s | No published figure | No published figure |
| M5 Max, 40-core GPU | 614 GB/s | About 120 tokens a second | About 3,220 tokens a second |
Bigger models are slower in proportion. A 70B model at 4-bit is more than ten times the size of that 7B test file. The one published 70B figure for a MacBook Pro comes from Chinese tech site ifanr, reported by 36Kr: Llama 3.3 70B (Q4_K_M) wrote 9.95 tokens a second on a 14-inch M5 Max with 128 GB. That is readable, but slower than you can skim.
Prompt reading decides how long you wait for the first word after pasting a long document. That is where Apple's Neural Accelerators help most.
M6 vs M5 Pro vs M5 Max: which chip is best for AI?
The M5 Max is the best chip for AI because it has the most memory and twice the M5 Pro's bandwidth. The M6 is the best value, but it stops at 32 GB. The M5 Pro sits in the middle: 64 GB for 70B models at half the M5 Max's speed.
| M6 | M5 Pro | M5 Max | |
|---|---|---|---|
| Most memory | 32 GB | 64 GB | 128 GB |
| Memory bandwidth | 153-170 GB/s | 307 GB/s | 460 or 614 GB/s |
| Biggest model at 4-bit (site rule) | 27B | 70B | About 120B; 70B at 8-bit |
| Cheapest 14-inch with most memory | £2,399 (32 GB) | £3,699 (64 GB) | £6,699 (128 GB) |
| Display options | Mini-LED only | Mini-LED or OLED | Mini-LED or OLED |
Choose the M6 if your models are 27B or smaller and you want the cheapest new MacBook Pro that handles them.
Choose the M5 Pro for 32B and 70B models on a budget. With 64 GB it loads Llama 3.3 70B, but nobody has published its 70B speed, and with half the M5 Max's bandwidth it will be slower.
Choose the M5 Max with the 40-core GPU if speed matters or you want 128 GB. Skip the 32-core M5 Max with 36 GB at £4,099 for AI: it gives a model about 27 GB, less than a 48 GB M5 Pro at £3,099. That matches the advice on the MacBook Pro M5 Pro and M5 Max page, which goes deeper on these two chips.
OLED or mini-LED: which MacBook Pro is better for local AI?
For local AI they are the same machine. The OLED and mini-LED models share the M5 Pro and M5 Max chips, the same memory options and the same bandwidth, so a model runs at the same speed on both. The OLED adds £400 to every configuration.
What the extra £400 buys is the screen and the body. The PROLED XDR panel reaches 2,000 nits peak and a 10,000,000:1 contrast ratio, against 1,600 nits and 1,000,000:1 on the mini-LED. It responds to touch. The 14-inch OLED with M5 Pro weighs 1.38 kg against 1.60 kg for the mini-LED, and Apple says its redesigned cooling keeps the same sustained performance in that lighter body.
Check that claim in reviews: Notebookcheck found the mini-LED 16-inch M5 Max (40-core GPU) settled at about 70 W under sustained load, and nobody had tested the OLED model by 29 October 2026. If the laptop is a tool for running models, the mini-LED version and £400 towards more memory is the better deal. If you also edit photos or video, the OLED screen is a real upgrade.
Can you still buy the M5 MacBook Pro, and for how much?
Yes, the M5 Pro and M5 Max mini-LED models are still on sale new from Apple, at the same UK prices as before the launch. The plain M5 14-inch is not: the M6 has taken its place in Apple's store.
Against the MacBook Pro M5 page (checked 17 October), the 14-inch M5 Pro is still £2,499 with 24 GB and £3,699 with 64 GB, and the 128 GB M5 Max still £6,699 (14-inch) or £6,999 (16-inch). Apple's refurbished store listed plain M5 14-inch models from £1,369 on 29 October; check the memory on each listing, because the plain M5 tops out at 32 GB.
Should you buy the new MacBook Pro now or wait?
Buy now if you need 64 GB or more in a laptop: no newer chip is on sale, and the OLED model uses the same M5 Pro and M5 Max. Wait a few weeks if you want the M6, because independent AI speed tests and Black Friday on 27 November are both close.
- M6, 32 GB: pre-order if you need it on 6 November. Otherwise wait for the first published AI speed tests and for retailer offers; Apple lists it as a standard model, so other shops can sell it.
- M5 Pro or M5 Max mini-LED: Prices did not fall when the OLED model arrived, and Apple has announced no M6 Pro or M6 Max, so I would not wait on a rumour.
- OLED: wait for reviews of sustained performance and battery life under load unless you want the screen for its own sake.
The software is free: Ollama and LM Studio both run on Apple silicon, and Ollama vs LM Studio explains which to install first.
Frequently asked questions
Is the MacBook Pro M6 good for local AI?
Yes, for models up to about 27B. With 32 GB at £2,399 it runs Qwen3.8 27B and similar models in Ollama or LM Studio. Its 170 GB/s memory is slow next to the M5 Pro and M5 Max, and it stops at 32 GB, so it cannot run 70B models.
Which MacBook Pro is best for running LLMs?
The 14-inch MacBook Pro with the 40-core M5 Max and 128 GB, at £6,699 for the mini-LED model as of 29 October 2026. It holds 70B models at 8-bit and has 614 GB/s of memory bandwidth. For 70B models at lower cost, the 14-inch M5 Pro with 64 GB costs £3,699.
Is the OLED MacBook Pro faster for AI?
No. The OLED and mini-LED MacBook Pro use the same M5 Pro and M5 Max chips with the same memory and bandwidth, so models run at the same speed. The OLED costs £400 more and adds a brighter touchscreen and a lighter body. Apple says sustained performance is the same.
How much RAM do I need in a MacBook Pro for AI?
16 GB runs 8B models, 24 GB runs 14B, 32 GB runs 27B, 48 GB runs 32B and 64 GB runs 70B models at 4-bit. 128 GB covers 70B at 8-bit and models up to about 120B. Memory cannot be upgraded later, so buy for the biggest model you expect to use.
Can the MacBook Pro M6 run a 70B model?
No. The M6 tops out at 32 GB, and Llama 3.3 70B needs about 42.4 GB at 4-bit before the conversation. You need at least 64 GB, which means the M5 Pro with the 20-core GPU at £3,699 or an M5 Max.
When can I get the new MacBook Pro in the UK?
Pre-orders for the M6 and OLED models opened on 28 October 2026. Apple says deliveries and in-store availability start on Friday 6 November 2026. The M5 Pro and M5 Max mini-LED models are available now at unchanged UK prices, from £2,499.
Read next
- MacBook Pro M5 Pro and M5 Max for local AI: 70B models on the move
- Which Mac or mini PC should you buy to run AI at home? (UK prices)
- Can my Mac run this AI model? Local LLM memory calculator
- Mac mini M6 for local AI: a quiet desktop for 8B-27B models
- Run a local LLM on your Mac: the complete guide to private AI agents
- Qwen 3.8 on a Mac: which model fits your memory
Sources
- MacBook Pro - Apple (UK) - checked 29 October 2026
- MacBook Pro with Liquid Retina XDR Display, Technical Specifications - Apple (UK) - checked 29 October 2026
- MacBook Pro with PROLED XDR Display, Technical Specifications - Apple (UK) - checked 29 October 2026
- MacBook Pro with Liquid Retina XDR Display, overview and performance footnotes - Apple (UK) - checked 29 October 2026
- MacBook Pro with PROLED XDR Display, overview - Apple (UK) - checked 29 October 2026
- Buy MacBook Pro (Liquid Retina XDR) - Apple UK store, every configuration price quoted - checked 29 October 2026
- Buy MacBook Pro with PROLED XDR Display - Apple UK store - checked 29 October 2026
- Apple UK education store, 14-inch MacBook Pro M6 32 GB - checked 29 October 2026
- Refurbished MacBook Pro - Apple (UK) - checked 29 October 2026
- Apple introduces the new 14-inch MacBook Pro with M6 - Apple Newsroom (UK), 27 October 2026 - checked 29 October 2026
- Apple unveils MacBook Pro with PROLED XDR Display - Apple Newsroom (UK), 28 October 2026 - checked 29 October 2026
- 9to5Mac, Apple announces special event for October 28 (press: the "Pro and Tell" name) - checked 29 October 2026
- Performance of llama.cpp on Apple Silicon M-series - llama.cpp Discussion #4167 - checked 29 October 2026
- 36Kr (ifanr), hands-on test of the M5 Max MacBook Pro with Llama 3.3 70B (press) - checked 29 October 2026
- Notebookcheck, Apple MacBook Pro 16 2026 M5 Max review (press) - checked 29 October 2026
