Argmax Pro SDK 3 Brings Qwen3-ASR to iOS with Word-Level Timestamps

solyarisoftware · x · 2026-08-01

Argmax released Pro SDK 3, successfully deploying the Qwen3-ASR-1.7b speech recognition model on iOS devices, enabling real-time transcription and custom vocabulary within 2GB of memory.

Developers also discovered that Qwen3-ASR can natively generate word-level timestamps without relying on an additional forced alignment model, drastically saving disk and memory overhead for on-device deployment.

Original post →

More from Models

Models channel →