Your rig

M4 Max · 48 GB · 48 GB

48 GB unified · 546 GB/s · 36.0 GB addressable by the GPU
Edit hardware Export as JSON

Context length

KV cache at f16, 32K tokens

Quality floor
Q8_0 — near-lossless Q6_K Q5_K_M Q4_K_M — recommended Q3_K_M Q2_K — damaged
KV cache type
Runtime
Ollamallama.cppLM StudioMLX

4 of 5 runtimes drive this device.

Allow CPU offload Include partial-offload fits

Offloaded layers run at 60 GB/s host bandwidth — expect single-digit tokens per second.

Which one should I actually run?

Best models for your M4 Max · 48 GB

Updated 8 Oct 2026 1 new this month

What do you want to do? Each row is the best-ranked model on that use case's shortlist that fits at Q5_K_M and 32K. The ordering is editorial; the sizes and speeds are computed. How the shortlists work.

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. Q5_K_M · 21.0 GB 17 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. Q5_K_M · 21.0 GB 17 Runs great
Best reasoning Gemma 4 31B
Google · Apache 2.0
89% AIME — the strongest maths of any model under 32B. Q5_K_M · 27.0 GB 15 Runs great
Best for writing Gemma 4 31B
Google · Apache 2.0
The cleanest, least verbose prose of the 2026 24–32 GB class. Q5_K_M · 27.0 GB 15 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. Q5_K_M · 21.0 GB 17 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). Q5_K_M · 21.0 GB 17 Runs great
Best translation Qwen3.8 27B
Alibaba · Apache 2.0
119 languages, and Qwen translations read as idiomatic rather than literal. Q5_K_M · 21.0 GB 17 Runs great
Fastest good model Qwen3.6 35B-A3B
Alibaba · Apache 2.0
3.3B active per token: the fastest model that still competes with dense 27Bs. Q5_K_M · 24.9 GB 65 Runs great
Best long context Qwen3.8 27B
Alibaba · Apache 2.0
262K native, and only 16 of 64 blocks keep a KV cache — long context is cheap here. Q5_K_M · 27.0 GB
at 128K context
17 Runs great

Everything that fits

Runs great

58 of 80 · fits with headroom for a longer prompt
ModelParamsQuantWeights +KV 32KTotal VRAM headroom Tok/sRuntime
Qwen3.6 35B-A3B 35.9B MoE / 3.3B act Q5_K_M 23.7 GB 0.63 GB 24.9 GB
11.1 GB free
65 MLX Run
Ornith 1.5 35B-A3B 35.9B MoE / 3.3B act Q5_K_M 23.7 GB 0.63 GB 24.9 GB
11.1 GB free
65 MLX Run
LLM-jp 4 33B Thinking 33.2B Q5_K_M 21.9 GB 8.00 GB 30.5 GB
5.5 GB free
14 MLX Run
Qwen3 32B 32.8B Q5_K_M 21.7 GB 8.00 GB 30.3 GB
5.7 GB free
14 MLX Run
DeepSeek-R1-Distill-Qwen 32B 32.8B Q5_K_M 21.7 GB 8.00 GB 30.3 GB
5.7 GB free
14 MLX Run
Qwen2.5-Coder 32B 32.8B Q5_K_M 21.7 GB 8.00 GB 30.3 GB
5.7 GB free
14 MLX Run
Olmo 3.1 32B Instruct 32.2B Q5_K_M 21.3 GB 2.75 GB 24.6 GB
11.4 GB free
14 MLX Run
Nemotron 3.5 Lightning 30B-A3B 31.6B MoE / 3.2B act Q5_K_M 20.9 GB 0.19 GB 21.6 GB
14.4 GB free
67 MLX Run
Gemma 4 31B 31.3B Q5_K_M 20.7 GB 5.78 GB 27.0 GB
9.0 GB free
15 MLX Run
GLM-4.7-Flash 30B-A3B 31.2B MoE / 3B act Q5_K_M 20.6 GB 1.65 GB 22.8 GB
13.2 GB free
72 MLX Run
Qwen3 30B-A3B 30.5B MoE / 3.3B act Q5_K_M 20.1 GB 3.00 GB 23.7 GB
12.3 GB free
65 MLX Run
Qwen3 Coder 30B-A3B 30.5B MoE / 3.3B act Q5_K_M 20.1 GB 3.00 GB 23.7 GB
12.3 GB free
65 MLX Run
Muse Glimmer 30B 29.8B Q5_K_M 19.7 GB 0.48 GB 20.8 GB
15.2 GB free
16 MLX Run
Granite 4.1 30B 28.9B Q5_K_M 19.1 GB 8.00 GB 27.7 GB
8.3 GB free
16 MLX Run
Qwen3.8 27B 27.8B Q5_K_M 18.4 GB 2.00 GB 21.0 GB
15.0 GB free
17 MLX Run
Qwen3.6 27B 27.8B Q5_K_M 18.4 GB 2.00 GB 21.0 GB
15.0 GB free
17 MLX Run
Gemma 3 27B 27.4B Q5_K_M 18.1 GB 2.91 GB 21.6 GB
14.4 GB free
17 MLX Run
Gemma 4 26B-A4B 26.5B MoE / 3.8B act Q5_K_M 17.5 GB 1.45 GB 19.5 GB
16.5 GB free
57 MLX Run
Devstral Small 2 24B 24B Q5_K_M 15.8 GB 5.00 GB 21.4 GB
14.6 GB free
19 MLX Run
Mistral Small 3.2 24B 23.6B Q5_K_M 15.6 GB 5.00 GB 21.2 GB
14.8 GB free
20 MLX Run
gpt-oss 20B 20.9B MoE / 3.6B act MXFP4 10.8 GB 0.75 GB 12.2 GB
23.8 GB free
76 MLX Run
Qwen3 14B 14.8B Q5_K_M 9.8 GB 5.00 GB 15.4 GB
20.6 GB free
31 MLX Run
DeepSeek-R1-Distill-Qwen 14B 14.8B Q5_K_M 9.8 GB 6.00 GB 16.4 GB
19.6 GB free
31 MLX Run
Qwen2.5-Coder 14B 14.8B Q5_K_M 9.8 GB 6.00 GB 16.4 GB
19.6 GB free
31 MLX Run
Phi-4 14B 14.7B Q5_K_M 9.7 GB 3.13 GB 13.4 GB
22.6 GB free
31 MLX Run
Ministral 3 14B 13.9B Q5_K_M 9.2 GB 5.00 GB 14.8 GB
21.2 GB free
33 MLX Run
Gemma 3 12B 12.2B Q5_K_M 8.1 GB 2.31 GB 11.0 GB
25.0 GB free
38 MLX Run
Mistral NeMo 12B 12.2B Q5_K_M 8.1 GB 5.00 GB 13.7 GB
22.3 GB free
38 MLX Run
Gemma 4 12B 12B Q5_K_M 7.9 GB 2.31 GB 10.8 GB
25.2 GB free
39 MLX Run
Qwen3.5 9B 9.65B Q5_K_M 6.4 GB 1.00 GB 8.0 GB
28.0 GB free
48 MLX Run
Ornith 1.5 9B 9.41B Q5_K_M 6.2 GB 1.00 GB 7.8 GB
28.2 GB free
49 MLX Run
Ministral 3 8B 8.92B Q5_K_M 5.9 GB 4.25 GB 10.7 GB
25.3 GB free
52 MLX Run
Granite 4.1 8B 8.79B Q5_K_M 5.8 GB 5.00 GB 11.4 GB
24.6 GB free
53 MLX Run
LFM2.5 8B-A1B 8.47B MoE / 1.5B act Q5_K_M 5.6 GB 0.38 GB 6.6 GB
29.4 GB free
144 MLX Run
Fara 7B 8.29B Q5_K_M 5.5 GB 1.75 GB 7.8 GB
28.2 GB free
56 MLX Run
Qwen3 8B 8.19B Q5_K_M 5.4 GB 4.50 GB 10.5 GB
25.5 GB free
56 MLX Run
Llama 3.1 8B Instruct 8.03B Q5_K_M 5.3 GB 4.00 GB 9.9 GB
26.1 GB free
58 MLX Run
Gemma 4 E4B 8.0B Q5_K_M 5.3 GB 0.47 GB 6.4 GB
29.6 GB free
58 MLX Run
Ling 3.0 Tiny 7.9B-A1.3B 7.9B MoE / 1.3B act Q5_K_M 5.2 GB 0.21 GB 6.0 GB
30.0 GB free
166 MLX Run
DeepSeek-R1-Distill-Qwen 7B 7.62B Q5_K_M 5.0 GB 1.75 GB 7.4 GB
28.6 GB free
61 MLX Run
Qwen2.5-Coder 7B 7.62B Q5_K_M 5.0 GB 1.75 GB 7.4 GB
28.6 GB free
61 MLX Run
Olmo 3 7B Instruct 7.3B Q5_K_M 4.8 GB 5.50 GB 10.9 GB
25.1 GB free
63 MLX Run
Mistral 7B Instruct v0.3 7.25B Q5_K_M 4.8 GB 4.00 GB 9.4 GB
26.6 GB free
64 MLX Run
Gemma 4 E2B 5.1B Q5_K_M 3.4 GB 0.23 GB 4.2 GB
31.8 GB free
91 MLX Run
Qwen3.5 4B 4.66B Q5_K_M 3.1 GB 1.00 GB 4.7 GB
31.3 GB free
99 MLX Run
Gemma 3 4B 4.3B Q5_K_M 2.8 GB 0.86 GB 4.3 GB
31.7 GB free
107 MLX Run
Qwen3 4B 4.02B Q5_K_M 2.7 GB 4.50 GB 7.8 GB
28.2 GB free
115 MLX Run
Ministral 3 3B 3.85B Q5_K_M 2.5 GB 3.25 GB 6.4 GB
29.6 GB free
120 MLX Run
Phi-4-mini 3.8B 3.84B Q5_K_M 2.5 GB 4.00 GB 7.1 GB
28.9 GB free
120 MLX Run
Granite 4.1 3B 3.4B Q5_K_M 2.2 GB 2.50 GB 5.3 GB
30.7 GB free
136 MLX Run
Llama 3.2 3B Instruct 3.21B Q5_K_M 2.1 GB 3.50 GB 6.2 GB
29.8 GB free
144 MLX Run
LFM2.5 2.6B 2.7B Q5_K_M 1.8 GB 0.50 GB 2.9 GB
33.1 GB free
171 MLX Run
Qwen3.5 2B 2.27B Q5_K_M 1.5 GB 0.38 GB 2.5 GB
33.5 GB free
204 MLX Run
Qwen3 1.7B 1.72B Q5_K_M 1.1 GB 3.50 GB 5.2 GB
30.8 GB free
269 MLX Run
Llama 3.2 1B Instruct 1.24B Q5_K_M 0.8 GB 1.00 GB 2.4 GB
33.6 GB free
373 MLX Run
Gemma 3 1B 1.0B Q5_K_M 0.7 GB 0.14 GB 1.4 GB
34.6 GB free
462 MLX Run
Qwen3.5 0.8B 0.87B Q5_K_M 0.6 GB 0.38 GB 1.5 GB
34.5 GB free
531 MLX Run
Qwen3 0.6B 0.6B Q5_K_M 0.4 GB 3.50 GB 4.5 GB
31.5 GB free
770 MLX Run

Won't fit in VRAM

22 models · some are possible with CPU offload
ModelParamsQuantWeights +KV 32KTotal Over budget Tok/sRuntime
Kimi K3 2.8T-A104B 2780B MoE / 104B act Q5_K_M 1835.0 GB 0.84 GB 1836.5 GB +1800.5 GB ~2.1 MLX Why
Qwen3.8 2.4T-A95B 2446B MoE / 95B act Q5_K_M 1614.5 GB 2.88 GB 1618.0 GB +1582.0 GB ~2.3 MLX Why
DeepSeek V4 Pro 1.6T-A49B 1650B MoE / 49B act Q5_K_M 1089.1 GB 2.14 GB 1091.9 GB +1055.9 GB ~4.4 MLX Why
Kimi K2.6 1T-A32B 1027B MoE / 32B act Q5_K_M 677.9 GB 2.14 GB 680.6 GB +644.6 GB ~6.7 MLX Why
GLM-5.2 744B-A40B 753B MoE / 40B act Q5_K_M 497.0 GB 2.74 GB 500.4 GB +464.4 GB ~5.4 MLX Why
GLM-5.3 744B-A40B 753B MoE / 40B act Q5_K_M 497.0 GB 2.74 GB 500.4 GB +464.4 GB ~5.4 MLX Why
DeepSeek-R1 671B 671B MoE / 37B act Q5_K_M 442.9 GB 2.14 GB 445.7 GB +409.7 GB ~5.8 MLX Why
DeepSeek V4.1 Flash 552B-A16B New 552B MoE / 16B act Q5_K_M 364.4 GB 0.15 GB 365.1 GB +329.1 GB ~13 DwarfStar Why
Qwen3.5 397B-A17B 403B MoE / 17B act Q5_K_M 266.0 GB 0.94 GB 267.5 GB +231.5 GB ~13 MLX Why
Ornith 1.5 397B-A17B 397B MoE / 17B act Q5_K_M 262.0 GB 0.94 GB 263.6 GB +227.6 GB ~13 MLX Why
GLM-5.3-Flash 320B-A18B 321B MoE / 18B act Q5_K_M 211.9 GB 0.34 GB 212.8 GB +176.8 GB ~12 MLX Why
DeepSeek V4 Flash 284B-A13B 284B MoE / 13B act Q5_K_M 187.5 GB 1.51 GB 189.6 GB +153.6 GB ~17 MLX Why
Qwen3 235B-A22B 235B MoE / 22B act Q5_K_M 155.1 GB 5.88 GB 161.6 GB +125.6 GB ~9.8 MLX Why
Qwen3.8-Flash-Next 180B-A6B 180B MoE / 6B act Q5_K_M 118.8 GB 0.75 GB 120.2 GB +84.2 GB ~36 MLX Why
Qwen3.5 122B-A10B 125B MoE / 10B act Q5_K_M 82.5 GB 0.75 GB 83.9 GB +47.9 GB ~22 MLX Why
Nemotron 3 Super 120B-A12B 124B MoE / 12B act Q5_K_M 81.8 GB 0.25 GB 82.7 GB +46.7 GB ~18 MLX Why
Ling 3.0 Flash 124B-A5B 124B MoE / 5.1B act Q5_K_M 81.8 GB 0.25 GB 82.7 GB +46.7 GB ~42 MLX Why
Mistral Small 4 119B-A6B 119B MoE / 6.5B act Q5_K_M 78.5 GB 0.70 GB 79.9 GB +43.9 GB ~33 MLX Why
gpt-oss 120B 117B MoE / 5.1B act MXFP4 60.5 GB 1.13 GB 62.2 GB +26.2 GB ~54 MLX Why
Llama 4 Scout 109B-A17B 109B MoE / 17B act Q5_K_M 71.9 GB 2.63 GB 75.2 GB +39.2 GB ~13 MLX Why
Llama 3.3 70B Instruct 70.6B Q5_K_M 46.6 GB 10.00 GB 57.2 GB +21.2 GB ~6.5 MLX Why
DeepSeek-R1-Distill-Llama 70B 70.6B Q5_K_M 46.6 GB 10.00 GB 57.2 GB +21.2 GB ~6.5 MLX Why

Totals = quantised weights + f16 KV cache at 32K tokens + 0.6 GB runtime overhead, against 36.0 GB usable unified memory. Tokens per second are estimated from 546 GB/s peak memory bandwidth at batch 1 — not measured. Read the method.


Same memory, different speed

These machines have the same 36.0 GB usable, so they run exactly the models above. Generation speed scales with memory bandwidth.

DeviceBandwidthSpeed vs yours
M4 Pro · 48 GB
M4 Pro · 2024
273 GB/s 0.50×

Best LLM for a 48 GB Mac, with every machine in the class →

What more memory unlocks

At Q4_K_M and 8K context this machine fits 58 of 80 models. Each step up, and the largest models it adds:

Every device ranked by what it can run →