Apple Silicon (Mac)
Apple M3 Pro
Data as of
- Memory bandwidth
- 150 GB/s
- Source
- Checked:
With 18 GB
One entry per model, up to 30, sorted by popularity. For each model we show the largest (most precise) quantization that fits in this memory with a context of at least 8K tokens. The context length is the longest one that quantization fits (for models whose maximum is under 8K, such as embedding models, that maximum). Models that are not for chat or text generation are labelled with their use. "Left" is the memory minus the estimate; under 5% of the memory is marked "tight" (not a guarantee that it fits).
-
Qwen3.8-27B UD-Q4_K_XLContext length 8,192 Estimated 19.3 GB 0.03 GB left (tight) Speed 5.1–6.8
-
Ornith-1.5-9B BF16Context length 8,192 Estimated 19.3 GB 0.04 GB left (tight) Speed 5.0–6.7
-
Context length 32,768 Estimated 18.9 GB 0.38 GB left (tight) Speed not estimated (MoE)
-
Context length 8,192 Estimated 19.1 GB 0.21 GB left (tight) Speed 5.2–6.9
-
Context length 32,768 Estimated 18.9 GB 0.47 GB left (tight) Speed not estimated (MoE)
-
Context length 8,192 Estimated 18.9 GB 0.46 GB left (tight) Speed 5.3–7.0
-
Context length 32,768 Estimated 19.0 GB 0.31 GB left (tight) Speed not estimated (MoE)
-
Context length 131,072 Estimated 17.4 GB 1.97 GB left Speed 6.6–8.8
-
Qwen3.8-27B-Uncensored Q4_K_MContext length 16,384 Estimated 19.1 GB 0.24 GB left (tight) Speed 5.4–7.1
-
Ornith-1.0-9B BF16Context length 8,192 Estimated 19.3 GB 0.01 GB left (tight) Speed 5.0–6.7
-
Context length 512 Estimated 2.41 GB 16.9 GB left Speed 67–90
-
Context length 32,768 Estimated 19.0 GB 0.32 GB left (tight) Speed not estimated (MoE)
-
Context length 131,072 Estimated 16.1 GB 3.19 GB left Speed 8.4–11
-
Context length 8,192 Estimated 18.4 GB 0.88 GB left (tight) Speed 5.5–7.3
-
Context length 32,768 Estimated 19.1 GB 0.27 GB left (tight) Speed 5.7–7.6
-
Context length 32,768 Estimated 19.3 GB 0.05 GB left (tight) Speed not estimated (MoE)
-
Context length 8,192 Estimated 18.4 GB 0.88 GB left (tight) Speed 5.5–7.3
-
Context length 8,192 Estimated 19.3 GB 0.04 GB left (tight) Speed 5.0–6.7
-
Context length 8,192 Estimated 19.3 GB 0.04 GB left (tight) Speed 5.0–6.7
-
Context length 8,192 Estimated 19.3 GB 0.01 GB left (tight) Speed not estimated (MoE)
-
Ternary-Bonsai-2-27B Q2_0Context length 131,072 Estimated 17.4 GB 1.91 GB left Speed 12–16
-
Context length 65,536 Estimated 17.6 GB 1.71 GB left Speed 7.4–9.9
-
Context length 262,144 Estimated 18.4 GB 0.97 GB left Speed 10–14
-
Context length 8,192 Estimated 19.3 GB 0.05 GB left (tight) Speed 5.1–6.8
-
Context length 8,192 Estimated 19.3 GB 0.03 GB left (tight) Speed 5.1–6.8
-
Wan2.1 Q4_K_MContext length 16,384 Estimated 19.1 GB 0.21 GB left (tight) Speed 5.4–7.1
-
Context length 65,536 Estimated 17.6 GB 1.71 GB left Speed 7.4–9.9
-
Context length 65,536 Estimated 12.5 GB 6.87 GB left Speed 22–29
-
Qwen3.8-35B-A3B-Distill Q3_K_MContext length 16,384 Estimated 19.1 GB 0.22 GB left (tight) Speed not estimated (MoE)
-
Context length 131,072 Estimated 19.2 GB 0.09 GB left (tight) Speed not estimated (MoE)
With 36 GB
One entry per model, up to 30, sorted by popularity. For each model we show the largest (most precise) quantization that fits in this memory with a context of at least 8K tokens. The context length is the longest one that quantization fits (for models whose maximum is under 8K, such as embedding models, that maximum). Models that are not for chat or text generation are labelled with their use. "Left" is the memory minus the estimate; under 5% of the memory is marked "tight" (not a guarantee that it fits).
-
Qwen3.8-27B UD-Q8_K_XLContext length 65,536 Estimated 37.0 GB 1.70 GB left (tight) Speed 2.9–3.8
-
Ornith-1.5-9B BF16Context length 262,144 Estimated 28.1 GB 10.5 GB left Speed 4.9–6.5
-
Context length 262,144 Estimated 35.7 GB 2.97 GB left Speed not estimated (MoE)
-
Context length 16,384 Estimated 38.3 GB 0.30 GB left (tight) Speed not estimated (MoE)
-
Context length 65,536 Estimated 37.0 GB 1.70 GB left (tight) Speed 2.9–3.8
-
Context length 262,144 Estimated 33.6 GB 5.04 GB left Speed not estimated (MoE)
-
Context length 262,144 Estimated 35.5 GB 3.15 GB left Speed 5.3–7.0
-
Context length 131,072 Estimated 36.0 GB 2.63 GB left Speed not estimated (MoE)
-
Context length 262,144 Estimated 29.7 GB 8.95 GB left Speed 3.8–5.0
-
Context length 65,536 Estimated 34.5 GB 4.11 GB left Speed 3.1–4.1
-
Ornith-1.0-9B BF16Context length 262,144 Estimated 27.6 GB 11.0 GB left Speed 5.0–6.7
-
Context length 512 Estimated 2.41 GB 36.2 GB left Speed 67–90
-
Context length 131,072 Estimated 36.4 GB 2.25 GB left Speed not estimated (MoE)
-
Context length 262,144 Estimated 28.1 GB 10.6 GB left Speed 4.9–6.5
-
Context length 40,960 Estimated 23.3 GB 15.4 GB left Speed 5.5–7.3
-
Context length 65,536 Estimated 37.0 GB 1.70 GB left (tight) Speed 2.9–3.8
-
Context length 262,144 Estimated 34.4 GB 4.26 GB left Speed not estimated (MoE)
-
Context length 131,072 Estimated 36.6 GB 2.09 GB left Speed 5.5–7.3
-
Context length 262,144 Estimated 28.1 GB 10.6 GB left Speed 4.9–6.5
-
Context length 262,144 Estimated 27.6 GB 11.0 GB left Speed 5.0–6.7
-
Context length 16,384 Estimated 38.3 GB 0.30 GB left (tight) Speed not estimated (MoE)
-
Ternary-Bonsai-2-27B Q2_0Context length 262,144 Estimated 26.0 GB 12.6 GB left Speed 12–16
-
Context length 262,144 Estimated 30.5 GB 8.15 GB left Speed 7.4–9.9
-
Context length 262,144 Estimated 18.4 GB 20.3 GB left Speed 10–14
-
Ornith-1.0-35B Q8_0Context length 16,384 Estimated 38.4 GB 0.28 GB left (tight) Speed not estimated (MoE)
-
Context length 65,536 Estimated 35.7 GB 2.92 GB left Speed 3.0–4.0
-
Context length 65,536 Estimated 37.0 GB 1.70 GB left (tight) Speed 2.9–3.8
-
Wan2.1 IQ4_NLContext length 131,072 Estimated 38.6 GB 0.03 GB left (tight) Speed 3.1–4.2
-
Context length 262,144 Estimated 30.5 GB 8.15 GB left Speed 7.4–9.9
-
Context length 65,536 Estimated 12.5 GB 26.2 GB left Speed 22–29
Frequently asked questions
- Which local LLMs run on Apple M3 Pro?
- On this site, 968 models have a quantization that fits in 36 GB at 8,192 tokens (estimated). Popular ones include Qwen3.8-27B, Ornith-1.5-9B, Huihui-Qwen3.6-35B-A3B-abliterated-MTP, Ornith-1.5-35B-A3B, Huihui-Qwen3.8-27B-abliterated.
- What is the largest model that runs on Apple M3 Pro?
- Among the models listed here, Mixtral-8x22B-Instruct-v0.1 (about 140.6B); IQ1_M fits in 36 GB (estimated at 8,192 tokens).
- How fast does Apple M3 Pro generate text?
- With 150 GB/s of memory bandwidth, Qwen3.8-27B at UD-Q8_K_XL (31.5 GB) runs at an estimated 2.9–3.8 tokens per second (60–80% of bandwidth ÷ file size; prompt processing not included).