Hardware / Apple /iPhone 15

Apple · Phone

iPhone 15

Runs 50 models fully in memory, 0 more with CPU offload. 1 of those sit close enough to 3.3 GB that our size estimate is what decides them.

Memory one app may use

3.3 GB of 6 GB

the most one app may use, not the memory in the phone — estimated

Bandwidth

51 GB/s

Runs

50

01

Specifications

Memory one app may use

3.3 GB of 6 GB

Bandwidth

51 GB/s

FP16 compute

—

TDP

—

MSRP

—

Street price

—

Released

Sep 22, 2023

Architecture

A16

02

What it runs

Best quant per model — sorted by fit score.

ModelParamsBest quantContextSpeedVerdict

Niagara 19m Batch.en20MQ4_K_M~262K1893 tok/sestFits in memory Niagara 38m Batch.en38MQ4_K_M~262K996 tok/sestFits in memory Audio8 ASR 0.1B0.3BQ4_K_M~66K126 tok/sestFits in memory Hubert Large Ls960 ft0.3BQ4_K_M~131K118 tok/sestFits in memory Wav2vec2 Large 960h Lv60 Self0.3BQ4_K_M~131K118 tok/sestFits in memory Canary 180m Flash0.2BQ4_K_M~131K210 tok/sestFits in memory Parakeet TDT CTC 110m0.1BQ4_K_M~262K344 tok/sestFits in memory Zipformer cr CTC Transducer XL 290M0.3BQ4_K_M~131K131 tok/sestFits in memory Zipformer Transducer XL 290M0.3BQ4_K_M~131K131 tok/sestFits in memory Moonshine Streaming Medium0.2BQ4_K_M~131K158 tok/sestFits in memory Moonshine Streaming Small0.1BQ4_K_M~66K379 tok/sestFits in memory Moonshine Streaming Tiny30MQ5_K_M~262K1076 tok/sestFits in memory Moonshine Tiny30MQ5_K_M~262K1076 tok/sestFits in memory ASR Conformer Largescaleasr0.5BQ4_K_M~131K79 tok/sestFits in memory Nemotron Speech Streaming en 0.6b0.6BQ4_K_M~66K63 tok/sestFits in memory Parakeet CTC 0.6b0.6BQ4_K_M~66K63 tok/sestFits in memory Parakeet RNNT 0.6b0.6BQ4_K_M~66K63 tok/sestFits in memory Parakeet TDT 0.6b v20.6BQ4_K_M~66K63 tok/sestFits in memory Parakeet TDT 0.6b v30.6BQ4_K_M~66K63 tok/sestFits in memory Nemotron 3.5 ASR Streaming 0.6b0.6BQ4_K_M~66K59 tok/sestFits in memory Distil Large v3.50.8BQ4_K_M~66K47 tok/sestFits in memory Whisper Large v3 Turbo0.8BQ4_K_M~66K47 tok/sestFits in memory Qwen3 ASR 0.6B0.8BQ4_K_M~66K49 tok/sestFits in memory Owsm CTC v3.1 1B1.1BQ4_K_M~66K34 tok/sestFits in memory Owsm CTC v3.2 ft 1B1BQ4_K_M~66K38 tok/sestFits in memory Owsm CTC v4 1B1BQ4_K_M~66K38 tok/sestFits in memory MMS 1b All1BQ4_K_M~66K38 tok/sestFits in memory Canary 1b1BQ4_K_M~66K38 tok/sestFits in memory Canary 1b Flash1BQ4_K_M~66K38 tok/sestFits in memory Canary 1b v21BQ4_K_M~66K38 tok/sestFits in memory Parakeet CTC 1.1b1.1BQ4_K_M~66K34 tok/sestFits in memory Parakeet RNNT 1.1b1.1BQ4_K_M~66K34 tok/sestFits in memory Parakeet TDT 1.1b1.1BQ4_K_M~66K34 tok/sestFits in memory ARK ASR 0.6B1.3BQ4_K_M~66K29 tok/sestFits in memory Llama 3.2 1B Instruct1.2BQ4_K_M33K32 tok/sestFits in memory Lite Whisper Large v3 Acc1.4BQ4_K_M~33K27 tok/sestFits in memory Nyra Health CrisperWhisper1.6BQ4_K_M~33K24 tok/sestFits in memory Whisper Large v31.5BQ4_K_M~33K25 tok/sestFits in memory Qwen3 ASR 0.6B HF0.8BQ4_K_M~8K47 tok/sestFits in memory MOSS Transcribe Diarize0.9BQ4_K_M~8K42 tok/sestFits in memory Cohere Transcribe 03 20262.1BQ4_K_M~16K18 tok/sestFits in memory Qwen3 ASR 1.7B2BQ4_K_M~16K19 tok/sestFits in memory Qwen3 ASR 1.7B HF2BQ4_K_M~16K19 tok/sestFits in memory STT 2.6b en2.6BQ4_K_M~8K15 tok/sestFits in memory Canary Qwen 2.5b2.6BQ4_K_M~8K15 tok/sestFits in memory GLM ASR Nano 25122.3BQ4_K_M~8K17 tok/sestFits in memory Granite 4.0 1b Speech2.3BQ4_K_M~4K17 tok/sestFits in memory Granite Speech 4.1 2b2.3BQ5_K_M~4K14 tok/sestFits in memory Granite Speech 4.1 2b NAR2.3BQ5_K_M~4K14 tok/sestFits in memory Higgs Audio v3 STT2.7BQ4_K_M~2K14 tok/sestFits in memoryest Laguna S 2.1118BQ4_K_M——Too large LongCat 2.01.8TQ8_0——Too large Inkling952BQ5_K_M——Too large Kimi K32.8TQ4_K_M——Too large Hy3299BQ8_0——Too large Nex-N2-Mini35.1BQ4_K_M——Too large GLM 5.2753BQ4_K_M——Too large Kimi K2.7 Code1.1TQ8_0——Too large Nemotron 3 Ultra561BQ5_K_M——Too large MiniMax M3427BQ8_0——Too large

A verdict marked est is a close call: the weight size behind it is calculated from the parameter count, and a 10% error would change it.

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