Hardware / Apple /Apple M3 (10-core GPU)

Apple · Apple Silicon

Apple M3 (10-core GPU)

for example MacBook Air (2024), MacBook Pro 14-inch (2023), iMac (2023)

Runs 94 models fully in memory, 22 more with CPU offload. 6 of those sit close enough to 24 GB that our size estimate is what decides them.

Memory

24 GB

Bandwidth

100 GB/s

Runs

94

01

Specifications

Memory

24 GB

Bandwidth

100 GB/s

FP16 compute

7.4 TFLOPS

TDP

22 W

MSRP

—

Street price

—

Released

Nov 7, 2023

Architecture

M3

02

What it runs

Best quant per model — sorted by fit score.

ModelParamsBest quantContextSpeedVerdict

Niagara 19m Batch.en20MQ4_K_M~262K5773 tok/sestFits in memory Niagara 38m Batch.en38MQ4_K_M~262K3039 tok/sestFits in memory Audio8 ASR 0.1B0.3BQ5_K_M~262K328 tok/sestFits in memory Distil Large v3.50.8BQ4_K_M~262K144 tok/sestFits in memory Hubert Large Ls960 ft0.3BQ4_K_M~262K361 tok/sestFits in memory Wav2vec2 Large 960h Lv60 Self0.3BQ5_K_M~262K308 tok/sestFits in memory Canary 180m Flash0.2BQ5_K_M~262K547 tok/sestFits in memory Nemotron 3.5 ASR Streaming 0.6b0.6BQ5_K_M~262K154 tok/sestFits in memory Nemotron Speech Streaming en 0.6b0.6BQ5_K_M~262K164 tok/sestFits in memory Parakeet CTC 0.6b0.6BQ5_K_M~262K164 tok/sestFits in memory Parakeet RNNT 0.6b0.6BQ5_K_M~262K164 tok/sestFits in memory Parakeet TDT CTC 110m0.1BQ8_0~262K599 tok/sestFits in memory Parakeet TDT 0.6b v20.6BQ4_K_M~262K192 tok/sestFits in memory Parakeet TDT 0.6b v30.6BQ5_K_M~262K164 tok/sestFits in memory Whisper Large v3 Turbo0.8BQ4_K_M~262K144 tok/sestFits in memory MOSS Transcribe Diarize0.9BQ4_K_M~131K128 tok/sestFits in memory Qwen3 ASR 0.6B0.8BQ5_K_M~262K126 tok/sestFits in memory Qwen3 ASR 0.6B HF0.8BQ4_K_M~131K144 tok/sestFits in memory Zipformer cr CTC Transducer XL 290M0.3BQ8_0~262K227 tok/sestFits in memory Zipformer Transducer XL 290M0.3BQ4_K_M~262K398 tok/sestFits in memory ASR Conformer Largescaleasr0.5BQ8_0~262K137 tok/sestFits in memory Moonshine Streaming Medium0.2BQ4_K_M~262K481 tok/sestFits in memory Moonshine Streaming Small0.1BQ8_0~262K659 tok/sestFits in memory Moonshine Streaming Tiny30MQ4_K_M~262K3849 tok/sestFits in memory Moonshine Tiny30MQ8_0~262K2196 tok/sestFits in memory Owsm CTC v3.2 ft 1B1BQ4_K_M~262K116 tok/sestFits in memory Owsm CTC v4 1B1BQ4_K_M~262K116 tok/sestFits in memory MMS 1b All1BQ4_K_M~262K116 tok/sestFits in memory Canary 1b1BQ4_K_M~262K116 tok/sestFits in memory Canary 1b Flash1BQ4_K_M~262K116 tok/sestFits in memory Canary 1b v21BQ4_K_M~262K116 tok/sestFits in memory Owsm CTC v3.1 1B1.1BQ4_K_M~262K105 tok/sestFits in memory Parakeet CTC 1.1b1.1BQ4_K_M~262K105 tok/sestFits in memory Parakeet RNNT 1.1b1.1BQ4_K_M~262K105 tok/sestFits in memory Parakeet TDT 1.1b1.1BQ4_K_M~262K105 tok/sestFits in memory ARK ASR 0.6B1.3BQ4_K_M~262K89 tok/sestFits in memory Llama 3.2 1B Instruct1.2BQ4_K_M33K96 tok/sestFits in memory Lite Whisper Large v3 Acc1.4BQ4_K_M~262K83 tok/sestFits in memory Whisper Large v31.5BQ4_K_M~262K77 tok/sestFits in memory Nyra Health CrisperWhisper1.6BQ4_K_M~262K72 tok/sestFits in memory Cohere Transcribe 03 20262.1BQ4_K_M~262K55 tok/sestFits in memory Qwen3 ASR 1.7B2BQ4_K_M~262K57 tok/sestFits in memory Qwen3 ASR 1.7B HF2BQ4_K_M~262K57 tok/sestFits in memory Granite 4.0 1b Speech2.3BQ4_K_M~131K50 tok/sestFits in memory Granite Speech 4.1 2b2.3BQ4_K_M~131K50 tok/sestFits in memory Granite Speech 4.1 2b NAR2.3BQ4_K_M~131K50 tok/sestFits in memory GLM ASR Nano 25122.3BQ4_K_M~262K50 tok/sestFits in memory Higgs Audio v3 STT2.7BQ4_K_M~131K43 tok/sestFits in memory STT 2.6b en2.6BQ4_K_M~262K44 tok/sestFits in memory Canary Qwen 2.5b2.6BQ4_K_M~262K44 tok/sestFits in memory Granite Speech 3.3 2b3BQ4_K_M~131K39 tok/sestFits in memory Llama 3.2 3B Instruct3.2BQ4_K_M131K36 tok/sestFits in memory MOSS Transcribe Preview 2B2.4BQ4_K_M~33K48 tok/sestFits in memory Ministral 3 3B 25123.8BQ4_K_M131K30 tok/sestFits in memory Granite 4.0 Micro3.2BQ4_K_M66K31 tok/sestFits in memory Gemma 3 4B4.3BQ4_K_M131K27 tok/sestFits in memory ARK ASR 3B4.1BQ4_K_M~262K28 tok/sestFits in memory Voxtral Mini 4B Realtime 26024.4BQ4_K_M~66K26 tok/sestFits in memory Hojo ASR V15.2BQ4_K_M~262K22 tok/sestFits in memory Voxtral Mini 3B 25075BQ4_K_M~262K23 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.

Something wrong on this page? Tell us