Core AI is Apple's on-device ML runtime in iOS 27 / macOS 27 and the successor to Core ML: PyTorch models are exported with Apple's coreai-torch (LLMs: coreai.llm.export) into .aimodel bundles that run on the GPU or the Neural Engine, e.g. Qwen3-8B 4-bit decodes at 94 tok/s on an M4 Max GPU, MLX 90 under the same protocol (apple-silicon-llm-bench, macOS 27 beta, 2026-06).

Parakeet-TDT-0.6B — Core AI

nvidia/parakeet-tdt-0.6b-v3 (cc-by-4.0, 600M) converted to Apple Core AI .aimodel — the first transducer / TDT (RNN-T family) ASR in the zoo. Transcribes ≤~29 s clips in 25 European languages as three stateless graphs + a host greedy loop (no LLM runtime).

• parakeet_encoder_float16_L2885.aimodel — FastConformer encoder + projector (fp16, ~1.2 GB), mel[1,128,2885] → enc_proj[1,361,640].
• parakeet_predict_float32.aimodel — embedding → 2-layer LSTM → projector (fp32), token[1,1],h,c[2,1,640] → dec_out[1,640],h',c'.
• parakeet_joint_float32.aimodel — head(relu(enc_frame+dec_out)) (fp32), → token_logits[1,8193], dur_logits[1,5].
• tokenizer.json (+ tokenizer_config.json), mel_filters_128x257_f32.bin (librosa-slaney).

Gated 77/77 token-exact end-to-end vs the HF ParakeetForTDT reference, and again token-exact through the Swift CoreAIKit KitParakeetModel. blank 8192 · durations [0,1,2,3,4] · 16 kHz.

Parakeet-TDT 0.6B on iPhone 17 Pro — the zoo's coreai-audio app, real speed.

Use it

⚡ One line — run the kit's task op on this model (import CoreAIOps; no session, no model plumbing, downloads on first use):

let text = try await CoreAI.transcribe(audioURL, options: .model("parakeet-tdt-0.6b-v3"))

Every op, one shape — Cookbook.

▶️ Run it (source) — the Transcribe runner (GUI + CLI, one app for every speech-to-text model in the catalog):

git clone https://github.com/john-rocky/coreai-kit
open coreai-kit/Examples/Transcribe/Transcribe.xcodeproj
# → Run, then pick "Parakeet-TDT 0.6B v3" in the model picker

# agents / headless (macOS):
cd coreai-kit/Examples/Transcribe
swift run transcribe-cli --model parakeet-tdt-0.6b-v3 --audio sample.wav

💻 Build with it — complete; the glue is kit API, copy-paste runs:

import CoreAIKit

let transcriber = try await KitTranscriber(catalog: "parakeet-tdt-0.6b-v3")
let samples = try AudioFile.pcm16kMono(url)  // any wav/m4a/mp3 → 16 kHz mono Float
let result = try await transcriber.transcribe(samples: samples)
// result.text (25 EU languages)

The take-home is Examples/Transcribe/Sources/QuickStart.swift — this exact code as one typed function, no UI; both the runner's GUI and its CLI call it. Recording? MicRecorder (kit API) captures mic audio as 16 kHz mono [Float] — the record button and permission prompt are your app's own chrome.

Integration checklist

• SPM: https://github.com/john-rocky/coreai-kit → product CoreAIKit
• Info.plist: NSMicrophoneUsageDescription — only if you record
• Entitlements: none needed (macOS)
• First run downloads the model — 1.3 GB (Mac) — then it loads from the local cache (Application Support; progress via the downloadProgress callback)
• Measure in Release — Debug is ~3× slower on per-token host work

Use (CoreAIKit)

let parakeet = try await KitParakeetModel(model: .parakeetTDT)
let result = try await parakeet.transcribe(samples: pcm16kMono)   // 16 kHz mono Float

---

More models in this format: Core AI Model Zoo — 75 models, each with the recipe that produced it.

Want a different model on-device? Open a request — free, open weights only; the export and its measured numbers get published publicly.