Close Menu
  • Home
  • Opinion
  • Region
    • Africa
    • Asia
    • Europe
    • Middle East
    • North America
    • Oceania
    • South America
  • AI & Machine Learning
  • Robotics & Automation
  • Space & Deep Tech
  • Web3 & Digital Economies
  • Climate & Sustainability Tech
  • Biotech & Future Health
  • Mobility & Smart Cities
  • Global Tech Pulse
  • Cybersecurity & Digital Rights
  • Future of Work & Education
  • Trend Radar & Startup Watch
  • Creator Economy & Culture
What's Hot

Why the 11-inch iPad Professional M5 Might Substitute Your Laptop computer

February 13, 2026

Eire has Europe’s largest digital abilities gender hole

February 13, 2026

OpenAI Releases a Analysis Preview of GPT‑5.3-Codex-Spark: A 15x Quicker AI Coding Mannequin Delivering Over 1000 Tokens Per Second on Cerebras {Hardware}

February 13, 2026
Facebook X (Twitter) Instagram LinkedIn RSS
NextTech NewsNextTech News
Facebook X (Twitter) Instagram LinkedIn RSS
  • Home
  • Africa
  • Asia
  • Europe
  • Middle East
  • North America
  • Oceania
  • South America
  • Opinion
Trending
  • Why the 11-inch iPad Professional M5 Might Substitute Your Laptop computer
  • Eire has Europe’s largest digital abilities gender hole
  • OpenAI Releases a Analysis Preview of GPT‑5.3-Codex-Spark: A 15x Quicker AI Coding Mannequin Delivering Over 1000 Tokens Per Second on Cerebras {Hardware}
  • Korea Bets on Ok-Manufacturers and Knowledge to Scale SME Exports By means of International Platforms – KoreaTechDesk
  • 8 Irish robotics start-ups it is best to learn about
  • Nicolas Cage Teased as Spider-Man in New Spider-Noir Trailer
  • AU Group Calls Capital Aid “A Strategic Crucial” for Banks following GTR MENA 2026 as Commerce Surges within the Area
  • UN-accredited org 1M1B’s push to empower climate-ready workforce
Friday, February 13
NextTech NewsNextTech News
Home - Asia - Alibaba Qwen Open-Sources Qwen3-ASR Speech Recognition Fashions: Supporting 52 Languages, with the 1.7B Model Reaching SOTA
Asia

Alibaba Qwen Open-Sources Qwen3-ASR Speech Recognition Fashions: Supporting 52 Languages, with the 1.7B Model Reaching SOTA

NextTechBy NextTechJanuary 31, 2026No Comments2 Mins Read
Share Facebook Twitter Pinterest LinkedIn Tumblr Telegram Email Copy Link
Follow Us
Google News Flipboard
Alibaba Qwen Open-Sources Qwen3-ASR Speech Recognition Fashions: Supporting 52 Languages, with the 1.7B Model Reaching SOTA
Share
Facebook Twitter LinkedIn Pinterest Email



January 29 — The Alibaba Qwen workforce has formally open-sourced the Qwen3-ASR mannequin collection, a robust lineup of speech recognition fashions developed beneath the Qwen household. The discharge contains two full-featured ASR fashions—Qwen3-ASR-1.7B and Qwen3-ASR-0.6B—in addition to an modern speech forced-alignment mannequin, Qwen3-ForcedAligner-0.6B. Collectively, the Qwen3-ASR collection helps speech recognition and language identification throughout 52 languages and dialects.

In response to Alibaba, Qwen3-ASR leverages a newly designed AuT pretrained speech encoder mixed with the robust multimodal basis of Qwen3-Omni, enabling extremely correct and secure speech recognition. The 1.7B mannequin achieves state-of-the-art (SOTA) efficiency throughout a number of eventualities, together with Mandarin Chinese language, English, Chinese language-accented speech, and singing voice recognition, whereas demonstrating robust robustness to advanced textual content and high-noise environments.

The 0.6B mannequin strikes a steadiness between efficiency and effectivity. Whereas sustaining excessive recognition accuracy, it helps 128-concurrent asynchronous inference with throughput as much as 2,000×, able to processing greater than 5 hours of audio in simply 10 seconds.

The Qwen3-ForcedAligner-0.6B is a timestamp prediction mannequin based mostly on non-autoregressive (NAR) giant language mannequin inference, supporting versatile and exact pressured alignment throughout 11 languages at arbitrary positions. Its timestamp accuracy surpasses conventional fashions resembling WhisperX and Nemo-Pressured-Aligner, attaining an environment friendly real-time issue (RTF) of 0.0089 beneath single-concurrency inference.

The Qwen workforce said that open-sourcing the Qwen3-ASR collection goals to speed up analysis and innovation in speech recognition and speech understanding. The mannequin architectures, weights, and a complete, user-friendly inference framework will all be launched as a part of the open-source bundle.

Supply: ITHome

Elevate your perspective with NextTech Information, the place innovation meets perception.
Uncover the most recent breakthroughs, get unique updates, and join with a world community of future-focused thinkers.
Unlock tomorrow’s traits as we speak: learn extra, subscribe to our publication, and develop into a part of the NextTech group at NextTech-news.com

Share. Facebook Twitter Pinterest LinkedIn Tumblr Email
NextTech
  • Website

Related Posts

Korea Bets on Ok-Manufacturers and Knowledge to Scale SME Exports By means of International Platforms – KoreaTechDesk

February 13, 2026

UN-accredited org 1M1B’s push to empower climate-ready workforce

February 12, 2026

The Hyperlink Between CX, EX, and Monitoring Intelligence

February 12, 2026
Add A Comment
Leave A Reply Cancel Reply

Economy News

Why the 11-inch iPad Professional M5 Might Substitute Your Laptop computer

By NextTechFebruary 13, 2026

Apple’s 11-inch iPad Professional M5, priced at $899 (was $999), is only a beast of…

Eire has Europe’s largest digital abilities gender hole

February 13, 2026

OpenAI Releases a Analysis Preview of GPT‑5.3-Codex-Spark: A 15x Quicker AI Coding Mannequin Delivering Over 1000 Tokens Per Second on Cerebras {Hardware}

February 13, 2026
Top Trending

Why the 11-inch iPad Professional M5 Might Substitute Your Laptop computer

By NextTechFebruary 13, 2026

Apple’s 11-inch iPad Professional M5, priced at $899 (was $999), is only…

Eire has Europe’s largest digital abilities gender hole

By NextTechFebruary 13, 2026

The report discovered that in an economic system near full employment, failing…

OpenAI Releases a Analysis Preview of GPT‑5.3-Codex-Spark: A 15x Quicker AI Coding Mannequin Delivering Over 1000 Tokens Per Second on Cerebras {Hardware}

By NextTechFebruary 13, 2026

OpenAI simply launched a brand new analysis preview referred to as GPT-5.3…

Subscribe to News

Get the latest sports news from NewsSite about world, sports and politics.

NEXTTECH-LOGO
Facebook X (Twitter) Instagram YouTube

AI & Machine Learning

Robotics & Automation

Space & Deep Tech

Web3 & Digital Economies

Climate & Sustainability Tech

Biotech & Future Health

Mobility & Smart Cities

Global Tech Pulse

Cybersecurity & Digital Rights

Future of Work & Education

Creator Economy & Culture

Trend Radar & Startup Watch

News By Region

Africa

Asia

Europe

Middle East

North America

Oceania

South America

2025 © NextTech-News. All Rights Reserved
  • About Us
  • Contact Us
  • Privacy Policy
  • Terms Of Service
  • Advertise With Us
  • Write For Us
  • Submit Article & Press Release

Type above and press Enter to search. Press Esc to cancel.

Subscribe For Latest Updates

Sign up to best of Tech news, informed analysis and opinions on what matters to you.

Invalid email address
 We respect your inbox and never send spam. You can unsubscribe from our newsletter at any time.     
Thanks for subscribing!