Best LLM for 32 GB unified memory · 2026

Best LLM for a 32 GB Mac: what fits, and what to run

Updated 8 Oct 2026 1 new this month

macOS lets the GPU wire down about 24.0 GB of a 32 GB Mac's unified memory; the rest stays with the system and your apps. At Q4_K_M with an 8K context, 58 of 80 models in the catalogue fit — 52 with headroom, 6 tightly. The pick for most people is Qwen3.8 27B: Best capability per gigabyte: a current-generation dense 27B with vision and 262K context. Speeds are estimated for an M4 · 32 GB; the fit verdicts are the same for every machine in the class.

Which one should I run?

One pick per job, from the editorial shortlist, checked to fit at Q4_K_M and 8K. Each label links the full ranking for that job. How picks are chosen.

RecommendationModelWhyQuant · totalTok/s est.
Best overall Qwen3.8 27B
Alibaba · Apache 2.0
Best capability per gigabyte: a current-generation dense 27B with vision and 262K context. Q4_K_M · 16.7 GB 4.3 Runs great
Best for coding Qwen3.8 27B
Alibaba · Apache 2.0
Terminal-Bench 73, DeepSWE 42 — a generation ahead of anything else that fits 24 GB. Q4_K_M · 16.7 GB 4.3 Runs great
Best reasoning Gemma 4 31B
Google · Apache 2.0
89% AIME — the strongest maths of any model under 32B. Q4_K_M · 20.2 GB 3.8 Runs great
Best for writing Gemma 4 31B
Google · Apache 2.0
The cleanest, least verbose prose of the 2026 24–32 GB class. Q4_K_M · 20.2 GB 3.8 Runs great
Best vision Qwen3.8 27B
Alibaba · Apache 2.0
Image and video in, OSWorld 84 — the strongest local vision-language model at 24 GB. Q4_K_M · 16.7 GB 4.3 Runs great
Best for agents Qwen3.8 27B
Alibaba · Apache 2.0
Computer use and tool calling are what 3.8 was trained for (OSWorld 84). Q4_K_M · 16.7 GB 4.3 Runs great
Best translation Qwen3.8 27B
Alibaba · Apache 2.0
119 languages, and Qwen translations read as idiomatic rather than literal. Q4_K_M · 16.7 GB 4.3 Runs great
Fastest good model Gemma 4 26B-A4B
Google · Apache 2.0
3.8B active and a small file — the quickest Gemma 4 with vision. Q4_K_M · 16.0 GB 15 Runs great
Best long context Qwen3.6 35B-A3B
Alibaba · Apache 2.0
262K with just 10 KV-cache layers and 2 KV heads: the cheapest long context at 24 GB. Q4_K_M · 23.3 GB
at 128K context
17 Tight fit

The long-context row is judged at 128K: Qwen3.6 35B-A3B holds that window on a 32 GB Mac at Q4_K_M with 2.50 GB of KV cache — that is why it can differ from the best overall pick.

Everything that fits a 32 GB Mac

Largest first. "Max context" is the longest window the Mac holds at that quantisation with an f16 cache; a q8_0 cache roughly doubles it.

ModelParamsQuantWeights+KV 8KTotalHeadroomMax contextTok/s est.
Olmo 3.1 32B Instruct 32.2B Q4_K_M 18.1 GB 1.25 GB 20.0 GB 4.0 GB 64K 3.7 Runs great
Nemotron 3.5 Lightning 30B-A3B 31.6B MoE Q4_K_M 17.8 GB 0.05 GB 18.4 GB 5.6 GB 256K 17 Runs great
Gemma 4 31B 31.3B Q4_K_M 17.6 GB 2.03 GB 20.2 GB 3.8 GB 32K 3.8 Runs great
GLM-4.7-Flash 30B-A3B 31.2B MoE Q4_K_M 17.5 GB 0.41 GB 18.6 GB 5.4 GB 113K 19 Runs great
Qwen3 30B-A3B 30.5B MoE Q4_K_M 17.1 GB 0.75 GB 18.5 GB 5.5 GB 66K 17 Runs great
Qwen3 Coder 30B-A3B 30.5B MoE Q4_K_M 17.1 GB 0.75 GB 18.5 GB 5.5 GB 66K 17 Runs great
Muse Glimmer 30B 29.8B Q4_K_M 16.8 GB 0.18 GB 17.5 GB 6.5 GB 128K 4.0 Runs great
Granite 4.1 30B 28.9B Q4_K_M 16.3 GB 2.00 GB 18.9 GB 5.1 GB 28K 4.1 Runs great
Qwen3.8 27B 27.8B Q4_K_M 15.6 GB 0.50 GB 16.7 GB 7.3 GB 124K 4.3 Runs great
Qwen3.6 27B 27.8B Q4_K_M 15.6 GB 0.50 GB 16.7 GB 7.3 GB 124K 4.3 Runs great
Gemma 3 27B 27.4B Q4_K_M 15.4 GB 1.03 GB 17.0 GB 7.0 GB 97K 4.4 Runs great
Gemma 4 26B-A4B 26.5B MoE Q4_K_M 14.9 GB 0.51 GB 16.0 GB 8.0 GB 212K 15 Runs great
Devstral Small 2 24B 24B Q4_K_M 13.5 GB 1.25 GB 15.3 GB 8.7 GB 63K 5.0 Runs great
Mistral Small 3.2 24B 23.6B Q4_K_M 13.3 GB 1.25 GB 15.1 GB 8.9 GB 64K 5.1 Runs great
gpt-oss 20B 20.9B MoE MXFP4 10.8 GB 0.19 GB 11.6 GB 12.4 GB 128K 17 Runs great
Qwen3 14B 14.8B Q4_K_M 8.3 GB 1.25 GB 10.2 GB 13.8 GB 96K 8.1 Runs great
DeepSeek-R1-Distill-Qwen 14B 14.8B Q4_K_M 8.3 GB 1.50 GB 10.4 GB 13.6 GB 80K 8.1 Runs great
Qwen2.5-Coder 14B 14.8B Q4_K_M 8.3 GB 1.50 GB 10.4 GB 13.6 GB 80K 8.1 Runs great
Phi-4 14B 14.7B Q4_K_M 8.3 GB 1.56 GB 10.4 GB 13.6 GB 16K 8.1 Runs great
Ministral 3 14B 13.9B Q4_K_M 7.8 GB 1.25 GB 9.7 GB 14.3 GB 99K 8.6 Runs great
Gemma 3 12B 12.2B Q4_K_M 6.9 GB 0.81 GB 8.3 GB 15.7 GB 128K 9.8 Runs great
Mistral NeMo 12B 12.2B Q4_K_M 6.9 GB 1.25 GB 8.7 GB 15.3 GB 105K 9.8 Runs great
Gemma 4 12B 12B Q4_K_M 6.7 GB 0.81 GB 8.2 GB 15.8 GB 256K 9.9 Runs great
Qwen3.5 9B 9.65B Q4_K_M 5.4 GB 0.25 GB 6.3 GB 17.7 GB 256K 12 Runs great
Ornith 1.5 9B 9.41B Q4_K_M 5.3 GB 0.25 GB 6.1 GB 17.9 GB 256K 13 Runs great
Ministral 3 8B 8.92B Q4_K_M 5.0 GB 1.06 GB 6.7 GB 17.3 GB 138K 13 Runs great
Granite 4.1 8B 8.79B Q4_K_M 4.9 GB 1.25 GB 6.8 GB 17.2 GB 118K 14 Runs great
LFM2.5 8B-A1B 8.47B MoE Q4_K_M 4.8 GB 0.09 GB 5.5 GB 18.5 GB 125K 37 Runs great
Fara 7B 8.29B Q4_K_M 4.7 GB 0.44 GB 5.7 GB 18.3 GB 125K 14 Runs great
Qwen3 8B 8.19B Q4_K_M 4.6 GB 1.13 GB 6.3 GB 17.7 GB 128K 15 Runs great
Llama 3.1 8B Instruct 8.03B Q4_K_M 4.5 GB 1.00 GB 6.1 GB 17.9 GB 128K 15 Runs great
Gemma 4 E4B 8.0B Q4_K_M 4.5 GB 0.14 GB 5.2 GB 18.8 GB 128K 15 Runs great
Ling 3.0 Tiny 7.9B-A1.3B 7.9B MoE Q4_K_M 4.4 GB 0.05 GB 5.1 GB 18.9 GB 128K 43 Runs great
DeepSeek-R1-Distill-Qwen 7B 7.62B Q4_K_M 4.3 GB 0.44 GB 5.3 GB 18.7 GB 128K 16 Runs great
Qwen2.5-Coder 7B 7.62B Q4_K_M 4.3 GB 0.44 GB 5.3 GB 18.7 GB 128K 16 Runs great
Olmo 3 7B Instruct 7.3B Q4_K_M 4.1 GB 2.50 GB 7.2 GB 16.8 GB 64K 16 Runs great
Mistral 7B Instruct v0.3 7.25B Q4_K_M 4.1 GB 1.00 GB 5.7 GB 18.3 GB 32K 16 Runs great
Gemma 4 E2B 5.1B Q4_K_M 2.9 GB 0.07 GB 3.5 GB 20.5 GB 128K 23 Runs great
Qwen3.5 4B 4.66B Q4_K_M 2.6 GB 0.25 GB 3.5 GB 20.5 GB 256K 26 Runs great
Gemma 3 4B 4.3B Q4_K_M 2.4 GB 0.30 GB 3.3 GB 20.7 GB 128K 28 Runs great
Qwen3 4B 4.02B Q4_K_M 2.3 GB 1.13 GB 4.0 GB 20.0 GB 32K 30 Runs great
Ministral 3 3B 3.85B Q4_K_M 2.2 GB 0.81 GB 3.6 GB 20.4 GB 209K 31 Runs great
Phi-4-mini 3.8B 3.84B Q4_K_M 2.2 GB 1.00 GB 3.8 GB 20.2 GB 128K 31 Runs great
Granite 4.1 3B 3.4B Q4_K_M 1.9 GB 0.63 GB 3.1 GB 20.9 GB 128K 35 Runs great
Llama 3.2 3B Instruct 3.21B Q4_K_M 1.8 GB 0.88 GB 3.3 GB 20.7 GB 128K 37 Runs great
LFM2.5 2.6B 2.7B Q4_K_M 1.5 GB 0.13 GB 2.2 GB 21.8 GB 128K 44 Runs great
Qwen3.5 2B 2.27B Q4_K_M 1.3 GB 0.09 GB 2.0 GB 22.0 GB 256K 53 Runs great
Qwen3 1.7B 1.72B Q4_K_M 1.0 GB 0.88 GB 2.4 GB 21.6 GB 32K 69 Runs great
Llama 3.2 1B Instruct 1.24B Q4_K_M 0.7 GB 0.25 GB 1.5 GB 22.5 GB 128K 96 Runs great
Gemma 3 1B 1.0B Q4_K_M 0.6 GB 0.04 GB 1.2 GB 22.8 GB 32K 119 Runs great
Qwen3.5 0.8B 0.87B Q4_K_M 0.5 GB 0.09 GB 1.2 GB 22.8 GB 256K 137 Runs great
Qwen3 0.6B 0.6B Q4_K_M 0.3 GB 0.88 GB 1.8 GB 22.2 GB 32K 199 Runs great
Qwen3.6 35B-A3B 35.9B MoE Q4_K_M 20.2 GB 0.16 GB 20.9 GB 3.1 GB 164K 17 Tight
Ornith 1.5 35B-A3B 35.9B MoE Q4_K_M 20.2 GB 0.16 GB 20.9 GB 3.1 GB 164K 17 Tight
LLM-jp 4 33B Thinking 33.2B Q4_K_M 18.7 GB 2.00 GB 21.3 GB 2.7 GB 18K 3.6 Tight
Qwen3 32B 32.8B Q4_K_M 18.4 GB 2.00 GB 21.0 GB 3.0 GB 19K 3.6 Tight
DeepSeek-R1-Distill-Qwen 32B 32.8B Q4_K_M 18.4 GB 2.00 GB 21.0 GB 3.0 GB 19K 3.6 Tight
Qwen2.5-Coder 32B 32.8B Q4_K_M 18.4 GB 2.00 GB 21.0 GB 3.0 GB 19K 3.6 Tight

Just out of reach

The smallest models that do not fit. Unified memory has no slower pool to spill into, so there is no offload fallback: each needs a bigger machine, or a lower quantisation.

ModelParamsTotal @ Q4_K_MOver budgetNeeds
Llama 3.3 70B Instruct 70.6B 42.8 GB +18.8 GB 96 GB Mac
DeepSeek-R1-Distill-Llama 70B 70.6B 42.8 GB +18.8 GB 96 GB Mac
gpt-oss 120B 117B MoE 61.4 GB +37.4 GB 128 GB Mac
Llama 4 Scout 109B-A17B 109B MoE 63.4 GB +39.4 GB 128 GB Mac
Mistral Small 4 119B-A6B 119B MoE 67.7 GB +43.7 GB 128 GB Mac
Ling 3.0 Flash 124B-A5B 124B MoE 70.4 GB +46.4 GB 128 GB Mac

Running it on a Mac

The GPU's share can be raised. The cap macOS puts on wired GPU memory is a setting, not a hardware limit: sudo sysctl iogpu.wired_limit_mb=… lifts it until the next reboot. Leave several gigabytes for macOS itself — every number on this page assumes the default share, so it under-promises rather than over.

MLX or GGUF. On Apple Silicon, MLX (via mlx-lm or LM Studio) is usually faster than llama.cpp's GGUF on the same model, and the community converts most popular releases within days. GGUF through Ollama or llama.cpp has the widest coverage and usually supports a new architecture first. The memory arithmetic is the same either way.

Speed is bandwidth. Every machine in this class holds the same models; what separates them is memory bandwidth (M1 Pro 200 GB/s, M2 Pro 200 GB/s, M4 120 GB/s), and tokens per second scales with it.


← 24 GB Mac All memory classes 36 GB Mac → Adjust context, quant or cache →

Every total is quantised weights + f16 KV cache at 8K + 0.6 GB runtime overhead, against 24.0 GB usable. Tokens per second are estimated from memory bandwidth and never measured. The method.