Which local LLMs can the Apple M4 Max (40-core GPU) run?

Unified memory — by default the GPU can use about 70% of it

Assumes 8k context and an f16 KV cache. Verdicts are shown for each memory size.

Specs

Memory options: 48 GB · 64 GB · 128 GB
Memory bandwidth: 546.0 GB/s
FP16 compute: 34.1 TFLOPS
Launch year: 2024

Verdicts at a glance

At Q4_K_M (or the closest available quant) with 8k context.

48 GBRuns great18 modelsRuns well6 modelsRuns slowly0 modelsWon't run5 models3 more models run at a lower quant.
64 GBRuns great18 modelsRuns well6 modelsRuns slowly1 modelWon't run4 models2 more models run at a lower quant.
128 GBRuns great20 modelsRuns well7 modelsRuns slowly1 modelWon't run1 model1 more model runs at a lower quant.

How to read the verdicts

Runs great: Fully on the GPU at 20 tok/s or more
Runs well: Fully on the GPU at 8–20 tok/s; MoE experts in system RAM at 20 tok/s or more; or at least 90% on the GPU at 8 tok/s or more
Runs slowly: 2–8 tok/s; CPU-only; less than 90% on the GPU; or MoE experts in system RAM below 20 tok/s
Won't run: Does not fit, or under 2 tok/s

Every model on the Apple M4 Max (40-core GPU)

Scroll sideways to see every column.

29 models sorted by verdict and speed, with 128 GB of memory
Model	Verdict	Speed	Quant	Memory	Runs as	Context	Notes
EXAONE 4.0 1.2B	Runs great	est. 249.4 tok/s199.6–299.3calibrated estimate ±20%	Q4_K_M	1.9 / 89.6 GB	Unified memory	up to 64k	—
HyperCLOVA X SEED 1.5B	Runs great	est. 180.5 tok/s144.4–216.6calibrated estimate ±20%	Q4_K_M	2.4 / 89.6 GB	Unified memory	up to 16k	—
gpt-oss-20b	Runs great	92.4 tok/s8k estimate 71.6 tok/s (57.3–85.9, calibrated estimate ±20%)measured (1 run, 2k context)	MXFP4	12.9 / 89.6 GB	Unified memory	up to 128k	reasoning model
Qwen3.5 35B-A3B	Runs great	est. 79.7 tok/s63.7–95.6calibrated estimate ±20%	Q4_K_M	23.4 / 89.6 GB	Unified memory	up to 256k	—
Qwen3 30B-A3B (2507)	Runs great	est. 58.2 tok/s46.6–69.9calibrated estimate ±20%	Q4_K_M	19.9 / 89.6 GB	Unified memory	up to 32k	—
Gemma 4 26B-A4B	Runs great	est. 56.7 tok/s45.3–68.0calibrated estimate ±20%	Q4_K_M	17.9 / 89.6 GB	Unified memory	up to 128k	—
DeepSeek R1 Distill Llama 8B	Runs great	est. 54.7 tok/s43.7–65.6calibrated estimate ±20%	Q4_K_M	6.6 / 89.6 GB	Unified memory	up to 32k	reasoning model
Kanana 1.5 8B	Runs great	est. 54.7 tok/s43.7–65.6calibrated estimate ±20%	Q4_K_M	6.6 / 89.6 GB	Unified memory	up to 32k	—
Llama 3.1 8B	Runs great	est. 54.7 tok/s43.7–65.6calibrated estimate ±20%	Q4_K_M	6.6 / 89.6 GB	Unified memory	up to 64k	—
Kanana 1.5 15.7B-A3B	Runs great	est. 53.4 tok/s42.7–64.1calibrated estimate ±20%	Q4_K_M	12.1 / 89.6 GB	Unified memory	up to 32k	—
Qwen3.5 9B	Runs great	est. 53.4 tok/s42.7–64.0calibrated estimate ±20%	Q4_K_M	6.7 / 89.6 GB	Unified memory	up to 256k	—
gpt-oss-120b	Runs great	est. 53.4 tok/s42.7–64.0calibrated estimate ±20%	MXFP4	64.3 / 89.6 GB	Unified memory	up to 64k	reasoning model
Qwen3 8B	Runs great	est. 52.5 tok/s42.0–63.0calibrated estimate ±20%	Q4_K_M	6.8 / 89.6 GB	Unified memory	up to 32k	—
Gemma 4 12B	Runs great	est. 42.8 tok/s34.2–51.3calibrated estimate ±20%	Q4_K_M	8.2 / 89.6 GB	Unified memory	up to 128k	—
Gemma 3 12B	Runs great	est. 41.8 tok/s33.4–50.2calibrated estimate ±20%	Q4_K_M	8.4 / 89.6 GB	Unified memory	up to 128k	—
HyperCLOVA X SEED Think 14B	Runs great	est. 32.1 tok/s22.5–41.8theoretical estimate ±30%	Q4_K_M	10.8 / 89.6 GB	Unified memory	up to 8k	reasoning model
Qwen3 14B	Runs great	est. 31.7 tok/s25.3–38.0calibrated estimate ±20%	Q4_K_M	10.9 / 89.6 GB	Unified memory	up to 32k	—
Phi-4	Runs great	est. 30.5 tok/s24.4–36.6calibrated estimate ±20%	Q4_K_M	11.3 / 89.6 GB	Unified memory	up to 8k	reasoning model
Qwen3.5 122B-A10B	Runs great	est. 25.4 tok/s20.3–30.4calibrated estimate ±20%	Q4_K_M	79.0 / 89.6 GB	Unified memory	up to 64k	—
Mistral Small 3.2 24B	Runs great	est. 20.9 tok/s16.7–25.1calibrated estimate ±20%	Q4_K_M	16.2 / 89.6 GB	Unified memory	up to 8k	—
Gemma 3 27B	Runs well	est. 19.0 tok/s15.2–22.8calibrated estimate ±20%	Q4_K_M	17.8 / 89.6 GB	Unified memory	up to 128k	Try IQ4_XS: Runs great
Qwen3.5 27B	Runs well	est. 18.6 tok/s14.9–22.3calibrated estimate ±20%	Q4_K_M	18.2 / 89.6 GB	Unified memory	up to 256k	Try IQ4_XS: Runs great
Solar Open 100B	Runs well	est. 18.4 tok/s14.7–22.1calibrated estimate ±20%	Q4_K_M	64.4 / 89.6 GB	Unified memory	up to 64k	Try IQ4_XS: Runs great
EXAONE 4.0 32B	Runs well	est. 16.5 tok/s13.2–19.8calibrated estimate ±20%	Q4_K_M	20.5 / 89.6 GB	Unified memory	up to 64k	reasoning model
EXAONE 4.5 33B	Runs well	est. 15.9 tok/s12.7–19.1calibrated estimate ±20%	Q4_K_M	21.2 / 89.6 GB	Unified memory	up to 64k	reasoning model
Qwen3 32B	Runs well	est. 15.0 tok/s12.0–17.9calibrated estimate ±20%	Q4_K_M	22.5 / 89.6 GB	Unified memory	up to 32k	Q2_K (heavy quality loss): Runs great
DeepSeek R1 Distill Qwen 32B	Runs well	est. 14.9 tok/s11.9–17.9calibrated estimate ±20%	Q4_K_M	22.6 / 89.6 GB	Unified memory	up to 16k	reasoning model
Llama 3.3 70B	Runs slowly	est. 7.2 tok/s5.8–8.7calibrated estimate ±20%	Q4_K_M	45.8 / 89.6 GB	Unified memory	up to 128k	Try IQ4_XS: Runs well
Solar Open 2 250B	Won't runTry Q2_K (heavy quality loss): Runs slowly	est. 26.8 tok/s18.7–34.8theoretical estimate ±30%	Q2_K	95.8 GB RAM	CPU only	up to 256k	CPU inference — capped at Runs slowly

Reasoning models spend extra tokens thinking, so their speed thresholds are 1.5× stricter (30 / 12 / 3 tok/s).

Verdict by memory size

Q4_K_M verdict for each memory configuration — open the model page for speeds
Model	48 GB	64 GB	128 GB
EXAONE 4.0 1.2B	Runs great	Runs great	Runs great
HyperCLOVA X SEED 1.5B	Runs great	Runs great	Runs great
gpt-oss-20b	Runs great	Runs great	Runs great
Qwen3.5 35B-A3B	Runs great	Runs great	Runs great
Qwen3 30B-A3B (2507)	Runs great	Runs great	Runs great
Gemma 4 26B-A4B	Runs great	Runs great	Runs great
DeepSeek R1 Distill Llama 8B	Runs great	Runs great	Runs great
Kanana 1.5 8B	Runs great	Runs great	Runs great
Llama 3.1 8B	Runs great	Runs great	Runs great
Kanana 1.5 15.7B-A3B	Runs great	Runs great	Runs great
Qwen3.5 9B	Runs great	Runs great	Runs great
gpt-oss-120b	Won't run	Won't run	Runs great
Qwen3 8B	Runs great	Runs great	Runs great
Gemma 4 12B	Runs great	Runs great	Runs great
Gemma 3 12B	Runs great	Runs great	Runs great
HyperCLOVA X SEED Think 14B	Runs great	Runs great	Runs great
Qwen3 14B	Runs great	Runs great	Runs great
Phi-4	Runs great	Runs great	Runs great
Qwen3.5 122B-A10B	Won't run	Won't run	Runs great
Mistral Small 3.2 24B	Runs great	Runs great	Runs great
Gemma 3 27B	Runs well	Runs well	Runs well
Qwen3.5 27B	Runs well	Runs well	Runs well
Solar Open 100B	Won't run	Won't run	Runs well
EXAONE 4.0 32B	Runs well	Runs well	Runs well
EXAONE 4.5 33B	Runs well	Runs well	Runs well
Qwen3 32B	Runs well	Runs well	Runs well
DeepSeek R1 Distill Qwen 32B	Runs well	Runs well	Runs well
Llama 3.3 70B	Won't run	Runs slowly	Runs slowly
Solar Open 2 250B	Won't run	Won't run	Won't run

Measured results on the Apple M4 Max (40-core GPU)

Public benchmarks we calibrate against. Their conditions (context, backend, flags) can differ from the estimates above.

Measured results on the Apple M4 Max (40-core GPU)
Model	Quant	Backend	Context	Prompt (tok/s)	Generation (tok/s)	Flags	Source	Measured
gpt-oss-20b	MXFP4	llama.cpp	2k	1,277.0	92.4	—	github.com	2025-08-15
llama-2-7b	Q4_0	llama.cpp	512	886.0	83.1	Metal	github.com	2024-11-15

Frequently asked questions

What is the largest model that runs entirely on the Apple M4 Max (40-core GPU)?

Qwen3.5 122B-A10B at Q4_K_M (a 78.26 GB file) fits entirely in 128 GB of unified memory with 8k context, at est. 25.4 tok/s (20.3–30.4, calibrated estimate ±20%).

How many local LLMs run well on the Apple M4 Max (40-core GPU)?

At Q4_K_M with 8k context, with 128 GB of memory, out of 29 tracked models: 20 run great, 7 run well, 1 runs slowly and 1 won't run.

Can the Apple M4 Max (40-core GPU) run a 70B model like Llama 3.3 70B?

Runs slowly — Unified memory, Q4_K_M: est. 7.2 tok/s (5.8–8.7, calibrated estimate ±20%).

Speeds are estimates from memory bandwidth, calibrated against public benchmarks, and each one comes with an error band and a confidence label. Real results vary with drivers, backend, context length and thermals.