What are the best models to run on Apple M2 Max (32GB)?

Based on model rankings and 24 GB usable VRAM: Top Overall: Qwen 3.8 27B (INT4) ~21 tok/s. Top Coding: Muse Glimmer 30B (INT4) ~19 tok/s. Optimal Daily Driver: Phi-4 Reasoning Plus (INT4) ~27 tok/s. Maximum Model Size: Falcon-40B (INT4) ~14 tok/s.

How are VRAM, throughput, and latency calculated?

Calculations are determined using apxml.com modeling, combining LLM architecture analysis with GPU hardware physics. VRAM and Context Scaling: Memory requirements are determined from model architecture, including parameter count, layer configuration, and attention structure. The KV cache footprint scales deterministically based on context length, batch size, and cache precision. Throughput (TPS): Token generation is calculated in the memory-bandwidth-bound decode phase, measuring the time required to stream model weights and KV cache relative to hardware memory bandwidth. Latency (TTFT): Time to first token is calculated from the compute operations required during prompt prefill relative to hardware tensor compute capacity.

How accurate are these estimates?

Calculations are calibrated against empirical ground-truth benchmark runs across diverse architectures and hardware. Because the underlying mechanics are benchmark-calibrated, estimates remain accurate across different configurations without needing to run physical tests for every scenario. VRAM Accuracy: Memory estimates are deterministic and closely match actual allocation. Minor differences typically stay within 3% to 5% due to CUDA runtime overhead or memory allocator fragmentation. Speed and Latency Accuracy: Throughput and TTFT typically vary within 5% to 10% depending on the specific inference runtime, kernel implementations, and system power limits.

Best LLMs to Run on Apple M2 Max (32GB)

Rank	Model	Parameters	Precision	TPS (Approx.)	TTFT
#36	Qwen 3.8 27B	27B	INT4	21 tok/s	~19.9s
#80	K2 Horizon MoVA 36B A4B	36B (4B active)	INT4	78 tok/s	~4.2s
#86	Qwen3.5-27B	27B	INT4	21 tok/s	~19.9s
#93	Sarvam-30B	32B (2.4B active)	INT4	106 tok/s	~2.7s
#99	Muse Glimmer 30B	30B	INT4	19 tok/s	~22.1s
#102	Phi-4 Reasoning Plus	14B	INT4 Q8	27 tok/s 18 tok/s	~10.5s ~10.5s
#103	Qwen3.6 35B A3B	35B (3B active)	INT4	91 tok/s	~3.4s
#105	MiMo V2 Flash	15B (309B active)	INT4 Q8	2 tok/s 1 tok/s	~217.7s ~237.1s
#110	Gemma 4 26B A4B	25.2B (3.8B active)	INT4	88 tok/s	~3.6s
#120	Qwen3.5-35B-A3B	35B (3B active)	INT4	91 tok/s	~3.4s
#121	Qwen3.5-9B	9B	INT4 Q8 FP16	53 tok/s 33 tok/s 17 tok/s	~6.8s ~6.8s ~7.4s
#127	K2 Horizon 7B	7B	INT4 Q8 FP16	65 tok/s 41 tok/s 22 tok/s	~5.3s ~5.3s ~5.3s
#130	Gemma 4 12B	11.95B	INT4 Q8	31 tok/s 21 tok/s	~9.0s ~9.0s
#136	Qwen3.5-4B	4B	INT4 Q8 FP16	97 tok/s 64 tok/s 36 tok/s	~3.1s ~3.1s ~3.1s
#139	GPT-OSS 20B	21B (3.6B active)	INT4	43 tok/s	~3.3s
#140	GLM-4V	9B	INT4 Q8 FP16	39 tok/s 27 tok/s 15 tok/s	~6.8s ~6.8s ~7.4s
#145	NVIDIA Nemotron 3 Nano 30B-A3B	3.5B (30B active)	INT4 Q8 FP16	18 tok/s 10 tok/s 5 tok/s	~21.4s ~21.4s ~21.5s
#150	Qwen3-30B-A3B	30B (3B active)	INT4	97 tok/s	~3.1s
#152	Granite 4.2 8B	8B	INT4 Q8 FP16	59 tok/s 36 tok/s 20 tok/s	~6.0s ~6.0s ~6.1s
#157	Gemma 4 E4B	8B	INT4 Q8 FP16	59 tok/s 36 tok/s 20 tok/s	~6.0s ~6.0s ~6.1s
#161	GLM-4	32B	INT4	12 tok/s	~27.7s
#163	Gemma 3 27B	27B	INT4	21 tok/s	~19.9s
#164	Phi-4	14B	INT4 Q8	37 tok/s 22 tok/s	~10.5s ~10.5s
#166	Qwen3-32B	32B	INT4	18 tok/s	~23.5s
#168	DeepSeek-R1 32B	32B	INT4	18 tok/s	~23.5s
#169	Gemma 4 31B	30.7B	INT4	18 tok/s	~22.6s
#171	Magistral Small	24B	INT4	17 tok/s	~17.7s
#172	Qwen2.5-32B	32B	INT4	18 tok/s	~23.5s
#174	OLMo 3.1 32B Instruct	32B	INT4	12 tok/s	~27.7s
#176	DeepSeek-R1 14B	14B	INT4 Q8	38 tok/s 22 tok/s	~10.5s ~10.5s
#177	Ministral 3 14B	14B	INT4 Q8	27 tok/s 18 tok/s	~10.5s ~10.5s
#180	Gemma 3 12B	12B	INT4 Q8 FP16	42 tok/s 25 tok/s 12 tok/s	~9.0s ~9.0s ~10.6s
#184	Mistral-Small-2501	24B	INT4	23 tok/s	~17.7s
#185	Ministral 3 8B	8B	INT4 Q8 FP16	43 tok/s 30 tok/s 17 tok/s	~6.0s ~6.0s ~6.1s
#186	Gemma 2 27B	27B	INT4	21 tok/s	~19.9s
#187	OLMo 3 7B Think	7B	INT4 Q8 FP16	47 tok/s 33 tok/s 20 tok/s	~5.3s ~5.3s ~5.3s
#188	Qwen3.5-2B	2B	INT4 Q8 FP16	142 tok/s 103 tok/s 64 tok/s	~1.6s ~1.7s ~1.7s
#189	MiniCPM5-2B	2B	INT4 Q8 FP16	142 tok/s 103 tok/s 64 tok/s	~1.6s ~1.7s ~1.7s
#190	Qwen2.5-14B	14B	INT4 Q8	37 tok/s 22 tok/s	~10.5s ~10.5s
#191	OLMo 3.1 32B Think	32B	INT4	12 tok/s	~27.7s
#193	Gemma 4 E2B	5.1B	INT4 Q8 FP16	82 tok/s 53 tok/s 29 tok/s	~3.9s ~3.9s ~3.9s
#195	Ministral 3 3B	3B	INT4 Q8 FP16	78 tok/s 61 tok/s 40 tok/s	~2.4s ~2.4s ~2.4s
#196	DeepSeek-R1 8B	8B	INT4 Q8 FP16	60 tok/s 37 tok/s 20 tok/s	~6.0s ~6.0s ~6.1s
#198	Gemma 2 9B	9B	INT4 Q8 FP16	53 tok/s 33 tok/s 17 tok/s	~6.8s ~6.8s ~7.4s
#199	Nemotron 3.5 Lightning 30B A3B	30B (3B active)	INT4	97 tok/s	~3.1s
#201	Qwen3-4B	4B	INT4 Q8 FP16	97 tok/s 64 tok/s 36 tok/s	~3.1s ~3.1s ~3.1s
#202	Qwen3-14B	14B	INT4 Q8	37 tok/s 22 tok/s	~10.5s ~10.5s
#203	Gemma 3 4B	4B	INT4 Q8 FP16	97 tok/s 64 tok/s 36 tok/s	~3.1s ~3.1s ~3.1s
#204	DeepSeek-R1 7B	7B	INT4 Q8 FP16	67 tok/s 41 tok/s 22 tok/s	~5.3s ~5.3s ~5.3s
#205	OLMo 3 32B Base	32B	INT4	12 tok/s	~27.7s
#206	Qwen3.5-0.8B	0.8B	INT4 Q8 FP16	208 tok/s 171 tok/s 122 tok/s	~670 ms ~672 ms ~674 ms
#207	Llama 3.1 8B	8B	INT4 Q8 FP16	59 tok/s 36 tok/s 20 tok/s	~6.0s ~6.0s ~6.1s
#207	Qwen3-8B	8B	INT4 Q8 FP16	59 tok/s 36 tok/s 20 tok/s	~6.0s ~6.0s ~6.1s
#210	Nemotron 3.5 Lightning	30B (3B active)	INT4	36 tok/s	~3.4s
#211	Command R	35B	INT4	11 tok/s	~30.2s
#212	Llama 3 8B	8B	INT4 Q8 FP16	59 tok/s 36 tok/s 20 tok/s	~6.0s ~6.0s ~6.1s
#213	Qwen3-1.7B	1.7B	INT4 Q8 FP16	154 tok/s 114 tok/s 73 tok/s	~1.4s ~1.4s ~1.4s
#214	Gemma 3 270M	0.27B	INT4 Q8 FP16	240 tok/s 222 tok/s 189 tok/s	~222 ms ~223 ms ~223 ms
#216	Qwen3-0.6B	0.6B	INT4 Q8 FP16	227 tok/s 192 tok/s 144 tok/s	~496 ms ~497 ms ~498 ms
#217	Phi-4-Mini	3.8B	INT4 Q8 FP16	100 tok/s 66 tok/s 38 tok/s	~3.0s ~3.0s ~3.0s
#218	Qwen2.5-7B	7B	INT4 Q8 FP16	65 tok/s 41 tok/s 22 tok/s	~5.3s ~5.3s ~5.3s
#220	Ministral-8B-2410	8B	INT4 Q8 FP16	59 tok/s 36 tok/s 20 tok/s	~6.0s ~6.0s ~6.1s
#221	DeepSeek-R1 1.5B	1.5B	INT4 Q8 FP16	163 tok/s 123 tok/s 80 tok/s	~1.2s ~1.3s ~1.3s
#222	Llama 3.2 3B	3B	INT4 Q8 FP16	115 tok/s 79 tok/s 47 tok/s	~2.4s ~2.4s ~2.4s
#223	Phi-3-mini	3.8B	INT4 Q8 FP16	100 tok/s 66 tok/s 38 tok/s	~3.0s ~3.0s ~3.0s
#224	Gemma 3n E2B IT	6B (2B active)	INT4 Q8 FP16	78 tok/s 67 tok/s 48 tok/s	~1.8s ~1.8s ~1.8s
#225	Gemma 2 2B	2B	INT4 Q8 FP16	142 tok/s 103 tok/s 64 tok/s	~1.6s ~1.7s ~1.7s
#226	Mistral-7B-Instruct-v0.1	7.3B	INT4 Q8 FP16	63 tok/s 39 tok/s 21 tok/s	~5.5s ~5.5s ~5.5s
#227	Llama 3.2 1B	1B	INT4 Q8 FP16	192 tok/s 153 tok/s 106 tok/s	~852 ms ~853 ms ~856 ms
#228	Phi-3-medium	14B	INT4 Q8	37 tok/s 22 tok/s	~10.5s ~10.5s
#229	Yi-34B	34B	INT4	12 tok/s	~29.4s
#230	OLMo 3 7B Instruct	7B	INT4 Q8 FP16	47 tok/s 33 tok/s 20 tok/s	~5.3s ~5.3s ~5.3s
#231	Qwen2-7B	7B	INT4 Q8 FP16	65 tok/s 41 tok/s 22 tok/s	~5.3s ~5.3s ~5.3s
#232	Phi-3-small	7B	INT4 Q8 FP16	65 tok/s 41 tok/s 22 tok/s	~5.3s ~5.3s ~5.3s
#233	Mistral-7B-Instruct-v0.2	7.3B	INT4 Q8 FP16	63 tok/s 39 tok/s 21 tok/s	~5.5s ~5.5s ~5.5s
#234	Gemma 3 1B	1B	INT4 Q8 FP16	192 tok/s 153 tok/s 106 tok/s	~852 ms ~853 ms ~856 ms
#236	Gemma 1 7B	7B	INT4 Q8 FP16	47 tok/s 33 tok/s 20 tok/s	~5.3s ~5.3s ~5.3s
#237	Gemma 1 2B	2B	INT4 Q8 FP16	151 tok/s 108 tok/s 66 tok/s	~1.6s ~1.7s ~1.7s
#238	ChatGLM3-6B	6B	INT4 Q8 FP16	52 tok/s 37 tok/s 23 tok/s	~4.6s ~4.6s ~4.6s
#239	ChatGLM2-6B	6B	INT4 Q8 FP16	52 tok/s 37 tok/s 23 tok/s	~4.6s ~4.6s ~4.6s
#240	ChatGLM-6B	6B	INT4 Q8 FP16	52 tok/s 37 tok/s 23 tok/s	~4.6s ~4.6s ~4.6s
Unranked Models (Sorted by Speed)
-	DINOv3 ViT-L/16	0.3B	INT4 Q8 FP16	260 tok/s 236 tok/s 195 tok/s	~247 ms ~247 ms ~248 ms
-	Qwen2-0.5B	0.5B	INT4 Q8 FP16	237 tok/s 205 tok/s 158 tok/s	~411 ms ~412 ms ~413 ms
-	Qwen2.5-0.5B	0.5B	INT4 Q8 FP16	237 tok/s 205 tok/s 158 tok/s	~411 ms ~412 ms ~413 ms
-	ERNIE-4.5-0.3B	0.3B	INT4 Q8 FP16	233 tok/s 215 tok/s 181 tok/s	~247 ms ~248 ms ~249 ms
-	ERNIE-4.5-0.3B-Base	0.3B	INT4 Q8 FP16	233 tok/s 215 tok/s 181 tok/s	~247 ms ~248 ms ~249 ms
-	Liquid D1-Omni 600M	0.6B	INT4 Q8 FP16	227 tok/s 192 tok/s 144 tok/s	~496 ms ~497 ms ~498 ms
-	Falcon-1B	1B	INT4 Q8 FP16	201 tok/s 158 tok/s 108 tok/s	~852 ms ~853 ms ~856 ms
-	Falcon3-1B	1B	INT4 Q8 FP16	192 tok/s 153 tok/s 106 tok/s	~852 ms ~853 ms ~856 ms
-	MiniCPM-V 4.7 1B	1B	INT4 Q8 FP16	192 tok/s 153 tok/s 106 tok/s	~852 ms ~853 ms ~856 ms
-	Qwen2-1.5B	1.5B	INT4 Q8 FP16	163 tok/s 123 tok/s 80 tok/s	~1.2s ~1.3s ~1.3s
-	Qwen2.5-1.5B	1.5B	INT4 Q8 FP16	163 tok/s 123 tok/s 80 tok/s	~1.2s ~1.3s ~1.3s
-	Spark X2.5 1.7B	1.7B	INT4 Q8 FP16	154 tok/s 114 tok/s 73 tok/s	~1.4s ~1.4s ~1.4s
-	Hy-MT2 1.8B	1.8B	INT4 Q8 FP16	150 tok/s 111 tok/s 70 tok/s	~1.5s ~1.5s ~1.5s
-	Falcon-3B	3B	INT4 Q8 FP16	121 tok/s 82 tok/s 47 tok/s	~2.4s ~2.4s ~2.4s
-	CroissantLLM Base	1.3B	INT4 Q8 FP16	119 tok/s 101 tok/s 73 tok/s	~1.1s ~1.1s ~1.1s
-	Phi-1	1.3B	INT4 Q8 FP16	119 tok/s 101 tok/s 73 tok/s	~1.1s ~1.1s ~1.1s
-	Phi-1.5	1.3B	INT4 Q8 FP16	119 tok/s 101 tok/s 73 tok/s	~1.1s ~1.1s ~1.1s
-	DeepSeek-R1 3B	3B	INT4 Q8 FP16	117 tok/s 80 tok/s 47 tok/s	~2.4s ~2.4s ~2.4s
-	Mellum 2.1 12B A2.5B Thinking	12B (2.5B active)	INT4 Q8 FP16	117 tok/s 85 tok/s 47 tok/s	~2.3s ~2.3s ~2.7s
-	Falcon3-3B	3B	INT4 Q8 FP16	115 tok/s 79 tok/s 47 tok/s	~2.4s ~2.4s ~2.4s
-	LFM2-VL 3B	3B	INT4 Q8 FP16	115 tok/s 79 tok/s 47 tok/s	~2.4s ~2.4s ~2.4s
-	Liquid D1 3B	3B	INT4 Q8 FP16	115 tok/s 79 tok/s 47 tok/s	~2.4s ~2.4s ~2.4s
-	Liquid D1-VL 3B	3B	INT4 Q8 FP16	115 tok/s 79 tok/s 47 tok/s	~2.4s ~2.4s ~2.4s
-	Liquid LFM2-VL 3B D1	3B	INT4 Q8 FP16	115 tok/s 79 tok/s 47 tok/s	~2.4s ~2.4s ~2.4s
-	Qwen2.5-3B	3B	INT4 Q8 FP16	115 tok/s 79 tok/s 47 tok/s	~2.4s ~2.4s ~2.4s
-	LFM2.5 24B A2B	24B (2.5B active)	INT4	108 tok/s	~2.6s
-	ERNIE-4.5-21B-A3B	21B (3B active)	INT4 Q8	101 tok/s 69 tok/s	~2.9s ~3.2s
-	ERNIE-4.5-21B-A3B-Base	21B (3B active)	INT4 Q8	101 tok/s 69 tok/s	~2.9s ~3.2s
-	ERNIE-4.5-VL-28B-A3B	28B (3B active)	INT4	98 tok/s	~3.1s
-	ERNIE-4.5-VL-28B-A3B-Base	28B (3B active)	INT4	98 tok/s	~3.1s
-	FrogNano 4B	4B	INT4 Q8 FP16	97 tok/s 64 tok/s 36 tok/s	~3.1s ~3.1s ~3.1s
-	Hy-Memory 30B A3B	30B (3B active)	INT4	97 tok/s	~3.1s
-	Hy-MT2 30B A3B	30B (3B active)	INT4	97 tok/s	~3.1s
-	Spark X2.5 4B	4B	INT4 Q8 FP16	97 tok/s 64 tok/s 36 tok/s	~3.1s ~3.1s ~3.1s
-	EmbeddingGemma 2	2B	INT4 Q8 FP16	93 tok/s 76 tok/s 53 tok/s	~1.7s ~1.7s ~1.7s
-	MiniCPM-V 4.7 35B A3B	35B (3B active)	INT4	91 tok/s	~3.4s
-	Phi-2	2.7B	INT4 Q8 FP16	82 tok/s 65 tok/s 43 tok/s	~2.2s ~2.2s ~2.2s
-	MaLLaM-3B	3B	INT4 Q8 FP16	78 tok/s 61 tok/s 40 tok/s	~2.4s ~2.4s ~2.4s
-	SmolLM3 3B	3B	INT4 Q8 FP16	78 tok/s 61 tok/s 40 tok/s	~2.4s ~2.4s ~2.4s
-	Falcon-7B	7B	INT4 Q8 FP16	68 tok/s 42 tok/s 22 tok/s	~5.3s ~5.3s ~5.3s
-	Falcon3-7B	7B	INT4 Q8 FP16	65 tok/s 41 tok/s 22 tok/s	~5.3s ~5.3s ~5.3s
-	Hy-MT2 7B	7B	INT4 Q8 FP16	65 tok/s 41 tok/s 22 tok/s	~5.3s ~5.3s ~5.3s
-	Mistral-7B-v0.1	7.3B	INT4 Q8 FP16	63 tok/s 39 tok/s 21 tok/s	~5.5s ~5.5s ~5.5s
-	Youtu Parsing Omni	8B	INT4 Q8 FP16	60 tok/s 37 tok/s 20 tok/s	~6.0s ~6.0s ~6.1s
-	LensVLM 9B	9B	INT4 Q8 FP16	53 tok/s 33 tok/s 17 tok/s	~6.8s ~6.8s ~7.4s
-	MiMo V2.6 Distill Qwen 9B	9B	INT4 Q8 FP16	53 tok/s 33 tok/s 17 tok/s	~6.8s ~6.8s ~7.4s
-	ChatGLM3-6B-32K	6B	INT4 Q8 FP16	52 tok/s 37 tok/s 23 tok/s	~4.6s ~4.6s ~4.6s
-	Yi-6B	6B	INT4 Q8 FP16	52 tok/s 37 tok/s 23 tok/s	~4.6s ~4.6s ~4.6s
-	Falcon3-10B	10B	INT4 Q8 FP16	49 tok/s 30 tok/s 15 tok/s	~7.5s ~7.5s ~8.2s
-	Kimi-VL-A3B-Instruct	16B (3B active)	INT4 Q8	49 tok/s 41 tok/s	~2.8s ~3.0s
-	Falcon2-11B	11B	INT4 Q8 FP16	47 tok/s 28 tok/s 14 tok/s	~8.2s ~8.3s ~9.0s
-	Hunyuan Lite	7B	INT4 Q8 FP16	47 tok/s 33 tok/s 20 tok/s	~5.3s ~5.3s ~5.3s
-	OLMo 3 7B Base	7B	INT4 Q8 FP16	47 tok/s 33 tok/s 20 tok/s	~5.3s ~5.3s ~5.3s
-	SEA-LION-7B	7.1B	INT4 Q8 FP16	46 tok/s 33 tok/s 19 tok/s	~5.4s ~5.4s ~5.4s
-	SEA-LION-7B-Instruct	7.1B	INT4 Q8 FP16	46 tok/s 33 tok/s 19 tok/s	~5.4s ~5.4s ~5.4s
-	Sahabat-AI-Llama3-8B-Instruct	8B	INT4 Q8 FP16	43 tok/s 30 tok/s 17 tok/s	~6.0s ~6.0s ~6.1s
-	Typhoon-2-8B	8B	INT4 Q8 FP16	43 tok/s 30 tok/s 17 tok/s	~6.0s ~6.0s ~6.1s
-	GLM-4-9B	9B	INT4 Q8 FP16	39 tok/s 27 tok/s 15 tok/s	~6.8s ~6.8s ~7.4s
-	GLM-4-9B-Chat	9B	INT4 Q8 FP16	39 tok/s 27 tok/s 15 tok/s	~6.8s ~6.8s ~7.4s
-	GLM-4-9B-Chat-1M	9B	INT4 Q8 FP16	39 tok/s 27 tok/s 15 tok/s	~6.8s ~6.8s ~7.4s
-	Yi-9B	9B	INT4 Q8 FP16	39 tok/s 27 tok/s 15 tok/s	~6.8s ~6.8s ~7.4s
-	Sahabat-AI-Gemma2-9B	9.2B	INT4 Q8 FP16	38 tok/s 26 tok/s 15 tok/s	~6.9s ~6.9s ~7.6s
-	Sahabat-AI-Gemma2-9B-Instruct	9.2B	INT4 Q8 FP16	38 tok/s 26 tok/s 15 tok/s	~6.9s ~6.9s ~7.6s
-	GreenMind-14B-R1	14B	INT4 Q8	27 tok/s 18 tok/s	~10.5s ~10.5s
-	AREX-2	27B	INT4	21 tok/s	~19.9s
-	PPLX Decider v1 27B	27B	INT4	21 tok/s	~19.9s
-	PPLX Decider v1.1 27B	27B	INT4	21 tok/s	~19.9s
-	Agnes 3.0 Flash	33B	INT4	17 tok/s	~24.2s
-	Falcon-40B	40B	INT4	14 tok/s	~31.8s

Apple M2 Max (32GB)

VRAM Scaling by Context Length

Best Models to Run on Apple M2 Max (32GB)

Inference Capacity

Fine-Tuning Capacity

Workload Recommendations

Frequently Asked Questions