PulseAugur
EN
LIVE 01:19:41
Deutsch(DE) Qwen/Qwen3-ASR-0.6B-hf ist ein Automatic-Speech-Recognition-Modell. Die Card nennt Sprachidentifikation und ASR fuer 52 Sprachen und Dialekte sowie Offline-/Str

Alibaba's Qwen releases two new multilingual ASR models

By PulseAugur Editorial · [2 sources] ·

Alibaba's Qwen team has released two new automatic speech recognition (ASR) models, Qwen3-ASR-1.7B-hf and Qwen3-ASR-0.6B-hf. Both models support 52 languages and dialects, with capabilities for offline and streaming inference. The larger 1.7B parameter model achieved a mean Word Error Rate (WER) of 5.59 on the Open ASR Leaderboard, while the smaller 0.6B model recorded a mean WER of 6.31. AI

IMPACT These models offer improved multilingual speech recognition capabilities, supporting both offline and streaming use cases.

RANK_REASON Release of new open-source models with benchmark results.

Read on Mastodon — mastodon.social →

AI-generated summary · Google Gemini · from 2 sources. How we write summaries →

Alibaba's Qwen releases two new multilingual ASR models

COVERAGE [2]

  1. Mastodon — mastodon.social TIER_1 Deutsch(DE) · aisyndicate ·

    Qwen/Qwen3-ASR-1.7B-hf is an Automatic Speech Recognition model. The card lists 52 languages and dialects, as well as offline and streaming inference. The README f

    Qwen/Qwen3-ASR-1.7B-hf ist ein Automatic-Speech-Recognition-Modell. Die Card nennt 52 Sprachen und Dialekte sowie Offline- und Streaming-Inferenz. Die README führt eine Mean WER von 5.59 auf dem Open ASR Leaderboard aus. https:// huggingface.co/Qwen/Qwen3-ASR- 1.7B-hf # KI # AI #…

    LINKS huggingface.co/…/Qwen3-ASR-1.7B-hf
  2. Mastodon — mastodon.social TIER_1 Deutsch(DE) · aisyndicate ·

    Qwen/Qwen3-ASR-0.6B-hf is an Automatic Speech Recognition model. The card mentions language identification and ASR for 52 languages and dialects as well as offline/str

    Qwen/Qwen3-ASR-0.6B-hf ist ein Automatic-Speech-Recognition-Modell. Die Card nennt Sprachidentifikation und ASR fuer 52 Sprachen und Dialekte sowie Offline-/Streaming-Inferenz. In den Benchmarks steht ein Mean WER von 6.31. https:// huggingface.co/Qwen/Qwen3-ASR- 0.6B-hf # KI # A…

    LINKS huggingface.co/…/Qwen3-ASR-0.6B-hf

RELATED ENTITIES

RELATED TOPICS