Your rig

M4 Pro · 24 GB · 24 GB

24 GB unified · 273 GB/s · 18.0 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 Pro · 24 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 Q4_K_M and 32K. The ordering is editorial; the sizes and speeds are computed. How the shortlists work.

RecommendationModelWhyQuant · totalTok/s est.
Best overall Gemma 4 12B
Google · Apache 2.0
The 12 GB generalist: text, image and audio in, 140+ languages. Q4_K_M · 9.7 GB 23 Runs great
Best for coding Qwen3.5 9B
Alibaba · Apache 2.0
The best coding model for 8–12 GB cards. Q4_K_M · 7.0 GB 28 Runs great
Best reasoning gpt-oss 20B
OpenAI · Apache 2.0
Low/medium/high reasoning effort in 13 GB. MXFP4 · 12.2 GB 38 Runs great
Best for writing Gemma 4 12B
Google · Apache 2.0
The 12 GB writing pick; 140+ languages. Q4_K_M · 9.7 GB 23 Runs great
Best vision Gemma 4 26B-A4B
Google · Apache 2.0
Vision at MoE speed in 18 GB. Q4_K_M · 16.9 GB 33 Tight fit
Best for agents Muse Glimmer 30B
Meta · Apache 2.0
Distilled specifically for always-on local agent workflows. Q4_K_M · 17.8 GB 9.1 Tight fit
Best translation Gemma 4 12B
Google · Apache 2.0
Broad language coverage on a 12 GB card. Q4_K_M · 9.7 GB 23 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 · 16.9 GB 33 Tight fit
Best long context Qwen3.5 9B
Alibaba · Apache 2.0
262K native in 8–12 GB. Q4_K_M · 10.0 GB
at 128K context
28 Runs great

Everything that fits

Runs great

38 of 80 · fits with headroom for a longer prompt
ModelParamsQuantWeights +KV 32KTotal VRAM headroom Tok/sRuntime
gpt-oss 20B 20.9B MoE / 3.6B act MXFP4 10.8 GB 0.75 GB 12.2 GB
5.8 GB free
38 MLX Run
Qwen3 14B 14.8B Q4_K_M 8.3 GB 5.00 GB 13.9 GB
4.1 GB free
18 MLX Run
DeepSeek-R1-Distill-Qwen 14B 14.8B Q4_K_M 8.3 GB 6.00 GB 14.9 GB
3.1 GB free
18 MLX Run
Qwen2.5-Coder 14B 14.8B Q4_K_M 8.3 GB 6.00 GB 14.9 GB
3.1 GB free
18 MLX Run
Phi-4 14B 14.7B Q4_K_M 8.3 GB 3.13 GB 12.0 GB
6.0 GB free
18 MLX Run
Ministral 3 14B 13.9B Q4_K_M 7.8 GB 5.00 GB 13.4 GB
4.6 GB free
20 MLX Run
Gemma 3 12B 12.2B Q4_K_M 6.9 GB 2.31 GB 9.8 GB
8.2 GB free
22 MLX Run
Mistral NeMo 12B 12.2B Q4_K_M 6.9 GB 5.00 GB 12.5 GB
5.5 GB free
22 MLX Run
Gemma 4 12B 12B Q4_K_M 6.7 GB 2.31 GB 9.7 GB
8.3 GB free
23 MLX Run
Qwen3.5 9B 9.65B Q4_K_M 5.4 GB 1.00 GB 7.0 GB
11.0 GB free
28 MLX Run
Ornith 1.5 9B 9.41B Q4_K_M 5.3 GB 1.00 GB 6.9 GB
11.1 GB free
29 MLX Run
Ministral 3 8B 8.92B Q4_K_M 5.0 GB 4.25 GB 9.9 GB
8.1 GB free
30 MLX Run
Granite 4.1 8B 8.79B Q4_K_M 4.9 GB 5.00 GB 10.5 GB
7.5 GB free
31 MLX Run
LFM2.5 8B-A1B 8.47B MoE / 1.5B act Q4_K_M 4.8 GB 0.38 GB 5.7 GB
12.3 GB free
84 MLX Run
Fara 7B 8.29B Q4_K_M 4.7 GB 1.75 GB 7.0 GB
11.0 GB free
33 MLX Run
Qwen3 8B 8.19B Q4_K_M 4.6 GB 4.50 GB 9.7 GB
8.3 GB free
33 MLX Run
Llama 3.1 8B Instruct 8.03B Q4_K_M 4.5 GB 4.00 GB 9.1 GB
8.9 GB free
34 MLX Run
Gemma 4 E4B 8.0B Q4_K_M 4.5 GB 0.47 GB 5.6 GB
12.4 GB free
34 MLX Run
Ling 3.0 Tiny 7.9B-A1.3B 7.9B MoE / 1.3B act Q4_K_M 4.4 GB 0.21 GB 5.3 GB
12.7 GB free
97 MLX Run
DeepSeek-R1-Distill-Qwen 7B 7.62B Q4_K_M 4.3 GB 1.75 GB 6.6 GB
11.4 GB free
36 MLX Run
Qwen2.5-Coder 7B 7.62B Q4_K_M 4.3 GB 1.75 GB 6.6 GB
11.4 GB free
36 MLX Run
Olmo 3 7B Instruct 7.3B Q4_K_M 4.1 GB 5.50 GB 10.2 GB
7.8 GB free
37 MLX Run
Mistral 7B Instruct v0.3 7.25B Q4_K_M 4.1 GB 4.00 GB 8.7 GB
9.3 GB free
37 MLX Run
Gemma 4 E2B 5.1B Q4_K_M 2.9 GB 0.23 GB 3.7 GB
14.3 GB free
53 MLX Run
Qwen3.5 4B 4.66B Q4_K_M 2.6 GB 1.00 GB 4.2 GB
13.8 GB free
58 MLX Run
Gemma 3 4B 4.3B Q4_K_M 2.4 GB 0.86 GB 3.9 GB
14.1 GB free
63 MLX Run
Qwen3 4B 4.02B Q4_K_M 2.3 GB 4.50 GB 7.4 GB
10.6 GB free
67 MLX Run
Ministral 3 3B 3.85B Q4_K_M 2.2 GB 3.25 GB 6.0 GB
12.0 GB free
70 MLX Run
Phi-4-mini 3.8B 3.84B Q4_K_M 2.2 GB 4.00 GB 6.8 GB
11.2 GB free
71 MLX Run
Granite 4.1 3B 3.4B Q4_K_M 1.9 GB 2.50 GB 5.0 GB
13.0 GB free
80 MLX Run
Llama 3.2 3B Instruct 3.21B Q4_K_M 1.8 GB 3.50 GB 5.9 GB
12.1 GB free
85 MLX Run
LFM2.5 2.6B 2.7B Q4_K_M 1.5 GB 0.50 GB 2.6 GB
15.4 GB free
100 MLX Run
Qwen3.5 2B 2.27B Q4_K_M 1.3 GB 0.38 GB 2.3 GB
15.7 GB free
120 MLX Run
Qwen3 1.7B 1.72B Q4_K_M 1.0 GB 3.50 GB 5.1 GB
12.9 GB free
158 MLX Run
Llama 3.2 1B Instruct 1.24B Q4_K_M 0.7 GB 1.00 GB 2.3 GB
15.7 GB free
219 MLX Run
Gemma 3 1B 1.0B Q4_K_M 0.6 GB 0.14 GB 1.3 GB
16.7 GB free
271 MLX Run
Qwen3.5 0.8B 0.87B Q4_K_M 0.5 GB 0.38 GB 1.5 GB
16.5 GB free
312 MLX Run
Qwen3 0.6B 0.6B Q4_K_M 0.3 GB 3.50 GB 4.4 GB
13.6 GB free
452 MLX Run

Runs — tight

2 models · fits at 32K, drop context or quant to go longer
ModelParamsQuantWeights +KV 32KTotal VRAM headroom Tok/sRuntime
Muse Glimmer 30B 29.8B Q4_K_M 16.8 GB 0.48 GB 17.8 GB
0.2 GB free
9.1 MLX Run
Gemma 4 26B-A4B 26.5B MoE / 3.8B act Q4_K_M 14.9 GB 1.45 GB 16.9 GB
1.1 GB free
33 MLX Run

Won't fit in VRAM

40 models · some are possible with CPU offload
ModelParamsQuantWeights +KV 32KTotal Over budget Tok/sRuntime
Kimi K3 2.8T-A104B 2780B MoE / 104B act Q4_K_M 1563.2 GB 0.84 GB 1564.6 GB +1546.6 GB ~1.2 MLX Why
Qwen3.8 2.4T-A95B 2446B MoE / 95B act Q4_K_M 1375.4 GB 2.88 GB 1378.8 GB +1360.8 GB ~1.3 MLX Why
DeepSeek V4 Pro 1.6T-A49B 1650B MoE / 49B act Q4_K_M 927.8 GB 2.14 GB 930.5 GB +912.5 GB ~2.6 MLX Why
Kimi K2.6 1T-A32B 1027B MoE / 32B act Q4_K_M 577.5 GB 2.14 GB 580.2 GB +562.2 GB ~4.0 MLX Why
GLM-5.2 744B-A40B 753B MoE / 40B act Q4_K_M 423.4 GB 2.74 GB 426.7 GB +408.7 GB ~3.2 MLX Why
GLM-5.3 744B-A40B 753B MoE / 40B act Q4_K_M 423.4 GB 2.74 GB 426.7 GB +408.7 GB ~3.2 MLX Why
DeepSeek-R1 671B 671B MoE / 37B act Q4_K_M 377.3 GB 2.14 GB 380.0 GB +362.0 GB ~3.4 MLX Why
DeepSeek V4.1 Flash 552B-A16B New 552B MoE / 16B act Q4_K_M 310.4 GB 0.15 GB 311.1 GB +293.1 GB ~7.9 DwarfStar Why
Qwen3.5 397B-A17B 403B MoE / 17B act Q4_K_M 226.6 GB 0.94 GB 228.1 GB +210.1 GB ~7.4 MLX Why
Ornith 1.5 397B-A17B 397B MoE / 17B act Q4_K_M 223.2 GB 0.94 GB 224.8 GB +206.8 GB ~7.4 MLX Why
GLM-5.3-Flash 320B-A18B 321B MoE / 18B act Q4_K_M 180.5 GB 0.34 GB 181.4 GB +163.4 GB ~7.0 MLX Why
DeepSeek V4 Flash 284B-A13B 284B MoE / 13B act Q4_K_M 159.7 GB 1.51 GB 161.8 GB +143.8 GB ~9.7 MLX Why
Qwen3 235B-A22B 235B MoE / 22B act Q4_K_M 132.1 GB 5.88 GB 138.6 GB +120.6 GB ~5.8 MLX Why
Qwen3.8-Flash-Next 180B-A6B 180B MoE / 6B act Q4_K_M 101.2 GB 0.75 GB 102.6 GB +84.6 GB ~21 MLX Why
Qwen3.5 122B-A10B 125B MoE / 10B act Q4_K_M 70.3 GB 0.75 GB 71.6 GB +53.6 GB ~13 MLX Why
Nemotron 3 Super 120B-A12B 124B MoE / 12B act Q4_K_M 69.7 GB 0.25 GB 70.6 GB +52.6 GB ~11 MLX Why
Ling 3.0 Flash 124B-A5B 124B MoE / 5.1B act Q4_K_M 69.7 GB 0.25 GB 70.6 GB +52.6 GB ~25 MLX Why
Mistral Small 4 119B-A6B 119B MoE / 6.5B act Q4_K_M 66.9 GB 0.70 GB 68.2 GB +50.2 GB ~19 MLX Why
gpt-oss 120B 117B MoE / 5.1B act MXFP4 60.5 GB 1.13 GB 62.2 GB +44.2 GB ~27 MLX Why
Llama 4 Scout 109B-A17B 109B MoE / 17B act Q4_K_M 61.3 GB 2.63 GB 64.5 GB +46.5 GB ~7.4 MLX Why
Llama 3.3 70B Instruct 70.6B Q4_K_M 39.7 GB 10.00 GB 50.3 GB +32.3 GB ~3.8 MLX Why
DeepSeek-R1-Distill-Llama 70B 70.6B Q4_K_M 39.7 GB 10.00 GB 50.3 GB +32.3 GB ~3.8 MLX Why
Qwen3.6 35B-A3B 35.9B MoE / 3.3B act Q4_K_M 20.2 GB 0.63 GB 21.4 GB +3.4 GB ~38 MLX Why
Ornith 1.5 35B-A3B 35.9B MoE / 3.3B act Q4_K_M 20.2 GB 0.63 GB 21.4 GB +3.4 GB ~38 MLX Why
LLM-jp 4 33B Thinking 33.2B Q4_K_M 18.7 GB 8.00 GB 27.3 GB +9.3 GB ~8.2 MLX Why
Qwen3 32B 32.8B Q4_K_M 18.4 GB 8.00 GB 27.0 GB +9.0 GB ~8.3 MLX Why
DeepSeek-R1-Distill-Qwen 32B 32.8B Q4_K_M 18.4 GB 8.00 GB 27.0 GB +9.0 GB ~8.3 MLX Why
Qwen2.5-Coder 32B 32.8B Q4_K_M 18.4 GB 8.00 GB 27.0 GB +9.0 GB ~8.3 MLX Why
Olmo 3.1 32B Instruct 32.2B Q4_K_M 18.1 GB 2.75 GB 21.5 GB +3.5 GB ~8.4 MLX Why
Nemotron 3.5 Lightning 30B-A3B 31.6B MoE / 3.2B act Q4_K_M 17.8 GB 0.19 GB 18.6 GB +0.6 GB ~40 MLX Why
Gemma 4 31B 31.3B Q4_K_M 17.6 GB 5.78 GB 24.0 GB +6.0 GB ~8.7 MLX Why
GLM-4.7-Flash 30B-A3B 31.2B MoE / 3B act Q4_K_M 17.5 GB 1.65 GB 19.8 GB +1.8 GB ~42 MLX Why
Qwen3 30B-A3B 30.5B MoE / 3.3B act Q4_K_M 17.1 GB 3.00 GB 20.7 GB +2.7 GB ~38 MLX Why
Qwen3 Coder 30B-A3B 30.5B MoE / 3.3B act Q4_K_M 17.1 GB 3.00 GB 20.7 GB +2.7 GB ~38 MLX Why
Granite 4.1 30B 28.9B Q4_K_M 16.3 GB 8.00 GB 24.9 GB +6.9 GB ~9.4 MLX Why
Qwen3.8 27B 27.8B Q4_K_M 15.6 GB 2.00 GB 18.2 GB +0.2 GB ~9.8 MLX Why
Qwen3.6 27B 27.8B Q4_K_M 15.6 GB 2.00 GB 18.2 GB +0.2 GB ~9.8 MLX Why
Gemma 3 27B 27.4B Q4_K_M 15.4 GB 2.91 GB 18.9 GB +0.9 GB ~9.9 MLX Why
Devstral Small 2 24B 24B Q4_K_M 13.5 GB 5.00 GB 19.1 GB +1.1 GB ~11 MLX Why
Mistral Small 3.2 24B 23.6B Q4_K_M 13.3 GB 5.00 GB 18.9 GB +0.9 GB ~11 MLX Why

Totals = quantised weights + f16 KV cache at 32K tokens + 0.6 GB runtime overhead, against 18.0 GB usable unified memory. Tokens per second are estimated from 273 GB/s peak memory bandwidth at batch 1 — not measured. Read the method.


Same memory, different speed

These machines have the same 18.0 GB usable, so they run exactly the models above. Generation speed scales with memory bandwidth.

DeviceBandwidthSpeed vs yours
M2 · 24 GB
M2 · 2022
100 GB/s 0.37×
M3 · 24 GB
M3 · 2023
100 GB/s 0.37×

Best LLM for a 24 GB Mac, with every machine in the class →

What more memory unlocks

At Q4_K_M and 8K context this machine fits 45 of 80 models. Each step up, and the largest models it adds:

Every device ranked by what it can run →