Hardware / Apple /Apple M4 (10-core GPU)
Apple · Apple Silicon
Apple M4 (10-core GPU)
for example MacBook Air (2025), MacBook Pro 14-inch (2024), Mac mini (2024), iMac (2024)
Runs 111 models fully in memory, 5 more with CPU offload. 4 of those sit close enough to 32 GB that our size estimate is what decides them.
Memory
32 GB
Bandwidth
120 GB/s
Runs
111
01
Specifications
Memory
32 GB
Bandwidth
120 GB/s
FP16 compute
8.5 TFLOPS
TDP
22 W
MSRP
—
Street price
—
Released
May 7, 2024
Architecture
M4
02
What it runs
Niagara 19m Batch.en20MQ8_0~262K3953 tok/sestFits in memoryNiagara 38m Batch.en38MQ5_K_M~262K3108 tok/sestFits in memoryAudio8 ASR 0.1B0.3BQ5_K_M~262K394 tok/sestFits in memoryDistil Large v3.50.8BQ4_K_M~262K173 tok/sestFits in memoryOwsm CTC v3.1 1B1.1BQ4_K_M~262K126 tok/sestFits in memoryOwsm CTC v3.2 ft 1B1BQ4_K_M~262K139 tok/sestFits in memoryOwsm CTC v4 1B1BQ5_K_M~262K118 tok/sestFits in memoryHubert Large Ls960 ft0.3BQ8_0~262K247 tok/sestFits in memoryMMS 1b All1BQ4_K_M~262K139 tok/sestFits in memoryWav2vec2 Large 960h Lv60 Self0.3BQ5_K_M~262K369 tok/sestFits in memoryCanary 180m Flash0.2BQ8_0~262K439 tok/sestFits in memoryCanary 1b1BQ4_K_M~262K139 tok/sestFits in memoryCanary 1b Flash1BQ4_K_M~262K139 tok/sestFits in memoryCanary 1b v21BQ5_K_M~262K118 tok/sestFits in memoryNemotron 3.5 ASR Streaming 0.6b0.6BQ8_0~262K124 tok/sestFits in memoryNemotron Speech Streaming en 0.6b0.6BQ8_0~262K132 tok/sestFits in memoryParakeet CTC 0.6b0.6BQ5_K_M~262K197 tok/sestFits in memoryParakeet CTC 1.1b1.1BQ4_K_M~262K126 tok/sestFits in memoryParakeet RNNT 0.6b0.6BQ4_K_M~262K231 tok/sestFits in memoryParakeet RNNT 1.1b1.1BQ4_K_M~262K126 tok/sestFits in memoryParakeet TDT CTC 110m0.1BQ8_0~262K719 tok/sestFits in memoryParakeet TDT 0.6b v20.6BQ5_K_M~262K197 tok/sestFits in memoryParakeet TDT 0.6b v30.6BQ8_0~262K132 tok/sestFits in memoryParakeet TDT 1.1b1.1BQ4_K_M~262K126 tok/sestFits in memoryWhisper Large v3 Turbo0.8BQ4_K_M~262K173 tok/sestFits in memoryMOSS Transcribe Diarize0.9BQ5_K_M~131K131 tok/sestFits in memoryQwen3 ASR 0.6B0.8BQ5_K_M~262K151 tok/sestFits in memoryQwen3 ASR 0.6B HF0.8BQ5_K_M~131K148 tok/sestFits in memoryZipformer cr CTC Transducer XL 290M0.3BQ5_K_M~262K407 tok/sestFits in memoryZipformer Transducer XL 290M0.3BQ5_K_M~262K407 tok/sestFits in memoryASR Conformer Largescaleasr0.5BQ4_K_M~262K289 tok/sestFits in memoryMoonshine Streaming Medium0.2BQ8_0~262K329 tok/sestFits in memoryMoonshine Streaming Small0.1BQ5_K_M~262K1181 tok/sestFits in memoryMoonshine Streaming Tiny30MQ5_K_M~262K3937 tok/sestFits in memoryMoonshine Tiny30MQ5_K_M~262K3937 tok/sestFits in memoryARK ASR 0.6B1.3BQ4_K_M~262K107 tok/sestFits in memoryLlama 3.2 1B Instruct1.2BQ4_K_M33K116 tok/sestFits in memoryLite Whisper Large v3 Acc1.4BQ4_K_M~262K99 tok/sestFits in memoryWhisper Large v31.5BQ4_K_M~262K92 tok/sestFits in memoryNyra Health CrisperWhisper1.6BQ4_K_M~262K87 tok/sestFits in memoryCohere Transcribe 03 20262.1BQ4_K_M~262K66 tok/sestFits in memoryQwen3 ASR 1.7B2BQ4_K_M~262K68 tok/sestFits in memoryQwen3 ASR 1.7B HF2BQ4_K_M~262K68 tok/sestFits in memoryGranite 4.0 1b Speech2.3BQ4_K_M~131K60 tok/sestFits in memoryGranite Speech 4.1 2b2.3BQ4_K_M~131K60 tok/sestFits in memoryGranite Speech 4.1 2b NAR2.3BQ4_K_M~131K60 tok/sestFits in memoryGLM ASR Nano 25122.3BQ4_K_M~262K60 tok/sestFits in memoryHiggs Audio v3 STT2.7BQ4_K_M~131K51 tok/sestFits in memorySTT 2.6b en2.6BQ4_K_M~262K53 tok/sestFits in memoryCanary Qwen 2.5b2.6BQ4_K_M~262K53 tok/sestFits in memoryGranite Speech 3.3 2b3BQ4_K_M~131K46 tok/sestFits in memoryLlama 3.2 3B Instruct3.2BQ4_K_M131K43 tok/sestFits in memoryMOSS Transcribe Preview 2B2.4BQ4_K_M~33K58 tok/sestFits in memoryMinistral 3 3B 25123.8BQ4_K_M131K37 tok/sestFits in memoryGranite 4.0 Micro3.2BQ4_K_M66K37 tok/sestFits in memoryGemma 3 4B4.3BQ4_K_M131K32 tok/sestFits in memoryARK ASR 3B4.1BQ4_K_M~262K34 tok/sestFits in memoryVoxtral Mini 4B Realtime 26024.4BQ4_K_M~131K32 tok/sestFits in memoryNemotron 3 Nano 30B A3B31.6BQ4_K_M33K39 tok/sestFits in memoryHojo ASR V15.2BQ4_K_M~262K27 tok/sestFits in memory
A speed marked est is calculated from this card's memory bandwidth and the size of the weights it has to read per word. It is not a measurement.