Skip to content

Mac Studio (M5 Max) AI benchmarks

Explore LLM, image and video benchmark results for Mac Studio (M5 Max), with hardware specifications and the software settings used in each test.

Hardware specifications

CPU
Apple M5 Max, 18-core CPU
GPU / SoC
Apple M5 Max, 40-core GPU
Memory
64 GB unified
GPU can use ~52 GB by default
INT8 compute
~74 TOPS
Memory bandwidth
614 GB/s
Storage
2 TB SSD
OS
macOS 27.0.1
Driver
Metal (built into macOS)

Benchmark results

ModelBenchmarkConfigurationMac Studio M5 MaxOlares One
Qwen3.8-27BIFBench
Evaluation time
Ollama (MLX), nvfp412.2 h8.7 h
Qwen3.8-27BGPQA Diamond
Evaluation time
Ollama (MLX), nvfp414.9 h11.5 h
Qwen3.8-27BInference speed
Token generation, single request
Ollama (MLX), nvfp474 tok/s93 tok/s
Qwen3.8-Flash-NextInference speed
Token generation, single request
Strata, Q2_0130 tok/s166 tok/s
Qwen-Image-2.1 BaseText to image
Per image
ComfyUI, int8, --gpu-only1.6 min12 s
Qwen-Image-2.1 BaseImage edit
Per image
ComfyUI, int8, --gpu-only1.9 min17 s
Qwen-Image-2.1 TurboImage edit
Per image
ComfyUI, int8, --gpu-only30 s5.2 s
MiniMax H3 FastH3 v2Text to video
Sampling time

720p, 5 s
ComfyUI, nvfp411 min78 s

Test configurations

Qwen3.8-27B

IFBench

Ollama (MLX), nvfp4

Engine and build
Ollama 0.40.15, MLX
Launch arguments
OLLAMA_CONTEXT_LENGTH=100352 OLLAMA_HOST=http://0.0.0.0:11434 OLLAMA_KEEP_ALIVE=30m0s OLLAMA_NUM_PARALLEL=1 OLLAMA_FLASH_ATTENTION=false

GPQA Diamond

Ollama (MLX), nvfp4

Engine and build
Ollama 0.40.15, MLX
Launch arguments
OLLAMA_CONTEXT_LENGTH=100352 OLLAMA_HOST=0.0.0.0:11434 OLLAMA_KEEP_ALIVE=30m OLLAMA_NUM_PARALLEL=1 OLLAMA_FLASH_ATTENTION=False

Inference speed

Ollama (MLX), nvfp4

Engine and build
Ollama 0.40.15, MLX
Launch arguments
OLLAMA_HOST=0.0.0.0:11434 OLLAMA_NUM_PARALLEL=4 OLLAMA_KEEP_ALIVE=30m OLLAMA_NO_CLOUD=0 OLLAMA_CONTEXT_LENGTH=100352

Inference speed

Ollama (GGML), UD-Q4_K_XL

Engine and build
Ollama 0.40.15, GGML/Metal (Ollama's llama.cpp runner, no speculative decoding)
Launch arguments
OLLAMA_HOST=0.0.0.0:11434 OLLAMA_NUM_PARALLEL=4 OLLAMA_KEEP_ALIVE=30m OLLAMA_NO_CLOUD=0 OLLAMA_CONTEXT_LENGTH=100352

Qwen3.8-Flash-Next

Inference speed

Strata, Q2_0

Engine and build
Strata strata 0.2.10 (engine 0.1.56)
Launch arguments
--pack q2_0 --native Qwen3.8-Flash-Next-GSQ-RCO-Q2_0-00001-of-00002.gguf --ple-gguf Qwen3.8-Flash-Next-GSQ-RCO-Q2_0-00002-of-00002.gguf --expert-profile expert-profile.bin --expert-cache 64 --prefill auto --spec 3 --spec-min-p 0.5 --mtp rt --max-context 131072 --mmap-experts --adapt-every 100000

Inference speed

Strata, IQ3_XXS, no MTP

Engine and build
Strata strata 0.2.10 (engine 0.1.56)
Launch arguments
--pack iq3_xxs --native Qwen3.8-Flash-Next-GSQ-RCO-IQ3_XXS-00001-of-00002.gguf --ple-gguf Qwen3.8-Flash-Next-GSQ-RCO-IQ3_XXS-00002-of-00002.gguf --expert-profile expert-profile.bin --expert-cache 64 --prefill auto --spec 2 --suffix-draft 0 --max-context 131072 --mmap-experts --adapt-every 100000

Inference speed

Strata, IQ3_S, no MTP

Engine and build
Strata strata 0.2.10 (engine 0.1.56)
Launch arguments
--pack iq3_s --native Qwen3.8-Flash-Next-GSQ-RCO-IQ3_S-00001-of-00002.gguf --ple-gguf Qwen3.8-Flash-Next-GSQ-RCO-IQ3_S-00002-of-00002.gguf --expert-profile expert-profile.bin --expert-cache 64 --prefill auto --spec 2 --suffix-draft 0 --max-context 131072 --mmap-experts --adapt-every 100000

Qwen-Image-2.1

ComfyUI, int8, --gpu-only

Qwen-Image-2.1 Base: Text to image, Image edit

Qwen-Image-2.1 Turbo: Image edit

Engine and build
ComfyUI v0.44.0-23-g63e5bf59, git checkout in comfyui, .venv (uv, python-build-standalone 3.12.16, PyTorch macOS wheels), MPS, torch 2.15.0, Python 3.12.16
Launch arguments
--listen 127.0.0.1 --port 8288 --extra-model-paths-config extra_model_paths.yaml --output-directory images --input-directory inputs --gpu-only

ComfyUI, int8

Qwen-Image-2.1 Base: Text to image, Image edit

Qwen-Image-2.1 Turbo: Image edit

Engine and build
ComfyUI v0.44.0-23-g63e5bf59, git checkout in comfyui, .venv (uv, python-build-standalone 3.12.16, PyTorch macOS wheels), MPS, torch 2.15.0, Python 3.12.16
Launch arguments
--listen 127.0.0.1 --port 8288 --extra-model-paths-config extra_model_paths.yaml --output-directory images --input-directory inputs

MiniMax H3

FastH3 v2: Text to video

ComfyUI, nvfp4

Engine and build
ComfyUI v0.44.0-23-g63e5bf59, git checkout in comfyui, .venv, MPS, torch 2.15.0, Python 3.12.16
Launch arguments
--listen 127.0.0.1 --port 8288 --disable-auto-launch --extra-model-paths-config extra_model_paths.yaml --output-directory outputs --gpu-only