Show HN: Which Mac runs which LLM, and how fast

A new MIT-licensed calculator published on GitHub and the benchmarks page estimates which large language models fit on which Apple Silicon Macs and how fast they generate, using two formulas: weights in gigabytes equal parameters in billions times bits per weight divided by eight, and speed equals memory bandwidth divided by weight size plus a fixed per-token cost. Fitted against eight dense runs on a base M4 and an M5 Pro, the model comes out at 90.8% of theoretical bandwidth plus 3.3 ms of overhead and reproduces every run to within 4.6%, predicting Llama 3.3 70B at Q4_K_M on an M4 Pro with 64 GB at roughly 5.7 tokens/sec with 42.8 GB of weights. The author cautions that the overhead term is held constant across chips, that prefill is compute bound and varies about 2x between model families (Qwen 2.5 7B at 1,130 prompt tokens/sec versus 587 for Llama 3.1 8B on identical hardware), and that every row outside the M4 and M5 Pro is calculated rather than measured.

Calculator Which Mac runs which LLM, and how fast Two numbers decide everything about running a model locally on Apple Silicon. Whether it fits in unified memory, which is arithmetic on the parameter count. And how fast it generates, which is memory bandwidth divided by the size of the weights. Pick your Mac and your model below and the calculator does both. Yes. Llama 3.3 70B at Q4 K M runs on M4 Pro with 64 GB. Expect roughly 5.7 tokens/sec of generation. The weights are 42.8 GB, the 8K tokens context adds about 1 GB, and 53.2 GB of your 64 GB is spoken for once macOS is accounted for. Rent a Mac that runs it: M4 Pro Mac mini with 64 GB $197/mo billed annually, or $286/mo billed monthly. A dedicated machine with Ollama and an OpenAI compatible endpoint, ready in about 5 minutes. That answer costs money. 56 GB of unified memory is not a cheap option on any Mac. Worth knowing before you commit to the hardware: a month on a rented M4 Pro with 64 GB costs $286, which is enough to run Llama 3.3 70B on your own documents and find out whether 5.7 tokens/sec is actually fast enough for what you had in mind. The two formulas Memory. Weights in gigabytes are parameters in billions times bits per weight divided by eight. Q4 K M averages about 4.85 bits once GGUF metadata is counted, Q8 0 about 8.5, FP16 exactly 16. Add 15% for runtime overhead, add the KV cache for your context length, then leave 3 GB for macOS. A model that fits with nothing spare will swap to SSD the moment you paste a long document, and swapping costs far more than any quantization ever saves you. Speed. Generating one token means reading every active weight out of memory exactly once, so bandwidth divided by weight size sets the ceiling. Sitting on top of that is a fixed cost per token that does not shrink with the model: sampling, kernel launches, framework overhead. Seconds per token is the sum of the two. Fitted against the eight dense runs we measured on two machines, a base M4 and an M5 Pro, that comes out at 90.8% of theoretical bandwidth plus 3.3 ms of overhead, and reproduces every one of those runs to within 4.6%. That second term is what most calculators leave out, and it is why their numbers stop making sense at the small end. Bandwidth alone predicts 525 tokens/sec for Llama 3.2 3B on an M5 Ultra. Ours predicts 177, because once the weights are down around two gigabytes the fixed cost per token, not the memory bus, is what you are up against. Method and raw JSON for the underlying measurements are on the benchmarks page https://macyou.co/benchmarks , and the model with its data and tests is on GitHub https://github.com/bagdaer1/mac-llm-calculator under MIT, so you can check it against your own runs. Where this math is wrong Worth knowing before you spend money on the strength of a number from a web page. - The overhead term is held constant across chips. In reality it almost certainly falls on faster silicon, which means the figures for Max and Ultra chips are more likely to be low than high. We would rather under promise a machine you are about to buy. - Prompt processing is a different number entirely. Prefill is compute bound, not bandwidth bound, and it varies by roughly 2x between model families at the same size. We measured Qwen 2.5 7B at 1,130 prompt tokens/sec against 587 for Llama 3.1 8B on identical hardware. If you feed the model whole documents and want short answers, that gap matters more than the generation figure everyone quotes. - Every row outside the M4 and the M5 Pro is calculated, not measured. Two machines is enough to show the formula travels: fitted on a base M4 at 120 GB/s, it predicts an M5 Pro at 307 GB/s to within 4.1% on dense models. It is not enough to call the rest measured. The rest matches what the community reports, and we label estimates as estimates. - Mixture of experts models were the thing we had wrong. Reading only the active experts predicted 91 tokens/sec for Qwen3 30B A3B and the machine did 78. Attention and the shared parts of the network are read for every token while only some experts are, so the real read is about 1.3x the active weight size. That correction is in the numbers above. It comes from one architecture and two models, so it is a correction with evidence behind it rather than a law. - Binned chips are slower than their name suggests. An M4 Max with a 32 core GPU runs at 410 GB/s and one with 40 cores at 546. Same name on the box, 33% difference in generation speed. Memory and speed for every model, at Q4 K M Minimum memory assumes an 8K context and 3 GB left for the system. Speeds are generation, in tokens per second. A dash means the model does not fit on that chip at its maximum memory configuration. | Model | Weights | Min memory | M4 | M4 Pro | M4 Max 40-core GPU | M5 Ultra | |---|---|---|---|---|---|---| | Llama 3.2 3B https://macyou.co/models/llama-3-2-3b | 1.9 GB | 8 GB | 47.5 | 90 | 140 | 199 | | Mistral 7B https://macyou.co/models/mistral-7b | 4.4 GB | 16 GB | 23.1 | 47.9 | 83 | 138 | | Qwen 2.5 7B https://macyou.co/models/qwen-2-5-7b | 4.6 GB | 16 GB | 21.9 | 45.8 | 80 | 134 | | Llama 3.1 8B https://macyou.co/models/llama-3-1-8b | 4.9 GB | 16 GB | 20.9 | 43.8 | 77 | 130 | | DeepSeek R1 Distill 8B https://macyou.co/models/deepseek-r1-8b | 4.9 GB | 16 GB | 20.9 | 43.8 | 77 | 130 | | Qwen 2.5 14B https://macyou.co/models/qwen-2-5-14b | 9 GB | 16 GB | 11.7 | 25.3 | 46.8 | 87 | | Qwen 2.5 32B https://macyou.co/models/qwen-2-5-32b | 19.9 GB | 32 GB | 5.4 | 12 | 23.1 | 46.5 | | Llama 3.3 70B https://macyou.co/models/llama-3-3-70b | 42.8 GB | 56 GB | — | 5.7 | 11.2 | 23.5 | | Qwen 2.5 72B https://macyou.co/models/qwen-2-5-72b | 44.1 GB | 56 GB | — | 5.5 | 10.8 | 22.9 | | Mistral Small 4 https://macyou.co/models/mistral-small-4 | 72.1 GB | 88 GB | — | — | 6.7 | 14.4 | | Mistral Large 2 https://macyou.co/models/mistral-large-2 | 74.6 GB | 96 GB | — | — | 6.5 | 13.9 | | Mistral Medium 3.5 https://macyou.co/models/mistral-medium-3-5 | 77.6 GB | 96 GB | — | — | 6.3 | 13.4 | | DeepSeek V4 Flash https://macyou.co/models/deepseek-v4-flash | 172.2 GB | 208 GB | — | — | — | 6.2 | | Llama 3.1 405B https://macyou.co/models/llama-3-1-405b | 245.5 GB | 288 GB | — | — | — | 4.4 | | DeepSeek V3 https://macyou.co/models/deepseek-v3 | 406.8 GB | 480 GB | — | — | — | 33.3 | Speeds for each chip assume its largest memory configuration: M4 at 32 GB, M4 Pro at 64 GB, M4 Max 40-core GPU at 128 GB, M5 Ultra at 512 GB. Apple Silicon memory bandwidth, M1 to M6 The single most useful table if you are choosing hardware, because bandwidth is what you are actually buying. Figures are Apple’s own published specifications, verified 2026-09-11. Each chip people commonly use for inference also has its own page https://macyou.co/local-llm with the full model by model breakdown at each memory configuration. | Chip | Bandwidth | Memory options | Found in | |---|---|---|---| | M1 | 68 GB/s | 8, 16 GB | MacBook Air, 13-inch MacBook Pro, Mac mini, iMac 2020-2021 16 GB ceiling makes 14B the hard limit, and only at Q4. | | M1 Pro | 200 GB/s | 16, 32 GB | MacBook Pro 14/16-inch 2021 | | M1 Max https://macyou.co/local-llm/m1-max | 400 GB/s | 32, 64 GB | MacBook Pro 14/16-inch 2021 , Mac Studio 2022 Still one of the best value-per-GB/s machines on the used market. | | M1 Ultra | 800 GB/s | 64, 128 GB | Mac Studio 2022 | | M2 | 100 GB/s | 8, 16, 24 GB | MacBook Air, 13-inch MacBook Pro, Mac mini 2022-2023 | | M2 Pro | 200 GB/s | 16, 32 GB | MacBook Pro 14/16-inch, Mac mini 2023 | | M2 Max | 400 GB/s | 32, 64, 96 GB | MacBook Pro 14/16-inch, Mac Studio 2023 | | M2 Ultra https://macyou.co/local-llm/m2-ultra | 800 GB/s | 64, 128, 192 GB | Mac Studio, Mac Pro 2023 192 GB at 800 GB/s still holds 123B-class models at Q8. | | M3 | 100 GB/s | 8, 16, 24 GB | MacBook Air, 14-inch MacBook Pro, iMac 2023-2024 | | M3 Pro | 150 GB/s | 18, 36 GB | MacBook Pro 14/16-inch 2023 Bandwidth dropped 25% versus M1 Pro and M2 Pro. The weakest Pro tier for inference. | | M3 Max 30-core GPU | 300 GB/s | 36, 96 GB | MacBook Pro 14/16-inch 2023 The binned M3 Max. 300 GB/s against 400 on the full chip. | | M3 Max 40-core GPU https://macyou.co/local-llm/m3-max | 400 GB/s | 48, 64, 128 GB | MacBook Pro 14/16-inch 2023 | | M3 Ultra https://macyou.co/local-llm/m3-ultra | 819 GB/s | 96, 256, 512 GB | Mac Studio 2025 512 GB was the first configuration able to hold a 671B MoE in memory. | | M4 https://macyou.co/local-llm/m4 | 120 GB/s | 16, 24, 32 GB | MacBook Air, 14-inch MacBook Pro, Mac mini, iMac 2024-2025 The one row on this page we have measured directly. See /benchmarks. | | M4 Pro https://macyou.co/local-llm/m4-pro | 273 GB/s | 24, 48, 64 GB | MacBook Pro 14/16-inch, Mac mini 2024 64 GB is the cheapest sensible way into 70B at Q4. | | M4 Max 32-core GPU | 410 GB/s | 36, 96 GB | MacBook Pro 14/16-inch 2024 , Mac Studio 2025 The binned M4 Max, 410 GB/s against 546 on the full chip. | | M4 Max 40-core GPU https://macyou.co/local-llm/m4-max | 546 GB/s | 48, 64, 128 GB | MacBook Pro 14/16-inch 2024 , Mac Studio 2025 | | M5 | 153 GB/s | 16, 24, 32 GB | MacBook Air, 14-inch MacBook Pro 2025-2026 | | M5 Pro https://macyou.co/local-llm/m5-pro | 307 GB/s | 24, 48, 64 GB | MacBook Pro 14/16-inch, Mac mini 2026 | | M5 Max base | 460 GB/s | 36, 48, 64, 128 GB | MacBook Pro 14/16-inch, Mac Studio 2026 Configurable to 614 GB/s. Check which one you are actually buying. | | M5 Max 614 GB/s https://macyou.co/local-llm/m5-max | 614 GB/s | 48, 64, 128 GB | MacBook Pro 14/16-inch, Mac Studio 2026 | | M5 Ultra https://macyou.co/local-llm/m5-ultra | 1200 GB/s | 96, 256, 512 GB | Mac Studio 2026 1.2 TB/s. The 512 GB tier exists only on the fully specced part, 36 core CPU and 80 core GPU; the configurations most people see stop at 256 GB. | | M6 https://macyou.co/local-llm/m6 | 153 GB/s | 16, 24, 32 GB | Mac mini 2026 Up to 170 GB/s on higher configurations; 153 is the entry figure used here. | Common questions How much RAM do I need to run a 70B model on a Mac? About 56 GB at Q4 K M with a short context: roughly 43 GB of weights, 15% runtime overhead, a gigabyte or two of KV cache, and 3 GB left for macOS. That makes 64 GB the smallest practical configuration, which in the current lineup means an M4 Pro, M5 Pro, or any Max or Ultra chip. At Q8 the same model needs around 128 GB. Does a faster GPU make local LLMs faster on a Mac? Barely, for generation. Producing one token means reading every active weight out of memory once, so generation speed is set by memory bandwidth, not GPU cores. A 24 GB M4 with more GPU cores runs an 8B model at the same speed as a 16 GB M4 because both sit on the same 120 GB/s bus. Prompt processing is the exception: that stage is compute bound and does scale with cores. How accurate are these speed estimates? They reproduce our own measurements to within 4.6%. The fit rests on ten runs across two machines: six models on a base M4 at 120 GB/s, and four on an M5 Pro at 307 GB/s. The second machine is the interesting one, because the constants were fitted before we had it. At two and a half times the bandwidth the dense predictions still land within 4.1%: 47.4 tokens/sec predicted for Qwen3 8B against 47.8 measured, 28.2 for Qwen 2.5 14B against 29.4. Everything outside those ten runs is extrapolation from the same two constants rather than a reading off a machine. Treat them as planning numbers, accurate enough to choose hardware by and not precise enough to quote. Why does the calculator reserve 3 GB for the system? Because a model that fits with nothing to spare is a model that swaps. macOS, a browser, and whatever agent process is calling the model all need memory, and once the machine starts paging to SSD the generation speed collapses regardless of what the bandwidth math says. Do mixture-of-experts models follow the same math? Memory yes, speed no. An MoE model has to hold every expert in unified memory, so the total parameter count decides whether it fits. Per token it reads only the active experts, so generation speed follows the active parameter count. That is why a 671B MoE with 37B active can generate faster than a dense 70B while needing eight times the memory. When the answer is a Mac you do not own If the model you want needs 64 GB or more and the machine has to be reachable around the clock, buying is not the only option. We rent dedicated Apple Silicon with Ollama already installed and an OpenAI compatible endpoint already exposed, from a base M4 Mac mini at $99/mo up to a 256 GB Mac Studio. Same hardware, same math as this page, none of the hosting.