Tencent launches Hy4 flash, cutting AI task costs by 95% in internal tests

Tencent's Hy4 flash promises to make AI-powered productivity tools more affordable, with internal tests showing a 95% cost reduction compared to its larger sibling.

Last Updated: November 16, 2026 Editorial Process
Editorial Process
See more of Inside AI's trusted news by adding us as a preferred source on Google.
AI neural network visualization
Published on: November 16, 2026

November 16, 2026, (Inside AI) — Tencent launched Hy4 flash on Monday, a lower-cost AI model for everyday work such as research, slide creation, and spreadsheet editing. In internal tests, the model cut AI task costs by 95% compared to the full Hy4 model. The API list prices are $0.07 per million input tokens, $0.281 per million output tokens, and $0.003 per million cached input tokens.

Hy4 flash supports multimodal inputs and a context window of up to one million tokens. It has 224 billion parameters and activates 12 billion per forward pass. On Tencent's internal WorkIndex benchmark, Hy4 flash cost one-twentieth as much per task as the full Hy4 model. The model is available in WorkBuddy and Yuanbao and through Tencent Cloud's TokenHub. WorkBuddy users can access it for free through November 30, and TokenHub is offering API access at half the list price. Tencent plans to release the model's weights and make it available on OpenRouter.

The launch underscores a growing industry focus on cost-efficient AI models that can handle routine tasks without the expense of larger models. Tencent's move follows similar efforts by other major AI labs to offer smaller, faster models for enterprise and consumer use.

Blind Test Shows Competitive Satisfaction

In a separate blind test, 85 WorkBuddy users rated outputs across 259 tasks. Tencent said the model scored close to Claude Haiku 5.5 in user satisfaction and had the lowest cost and shortest completion time among the models tested. The test results suggest that Hy4 flash can deliver comparable user experience at a fraction of the cost.

Read: Japan's Slow AI Adoption

The model's architecture, with 224 billion total parameters but only 12 billion active per forward pass, reflects a mixture-of-experts design that balances capability and efficiency. This approach allows the model to handle complex tasks while keeping inference costs low.

Tencent's decision to release the model's weights and make it available on OpenRouter could further drive adoption among developers and researchers. OpenRouter is a platform that provides access to various AI models through a unified API.

The free access for WorkBuddy users through November 30 and the half-price API offer on TokenHub are likely aimed at attracting users and generating feedback. These promotional efforts may help Tencent compete with other AI providers in the crowded market for AI-powered productivity tools.

Hy4 flash's multimodal capabilities and large context window make it suitable for tasks that involve both text and images, such as creating presentations or analyzing documents. The one-million-token context window allows the model to process lengthy inputs, which is useful for research and data analysis.

The launch of Hy4 flash is part of Tencent's broader strategy to integrate AI into its products and services. The company has been investing heavily in AI research and development, and this model represents a step toward making AI more accessible and affordable for everyday use.

As AI models become more commoditized, cost efficiency and user satisfaction will be key differentiators. Tencent's internal tests show that Hy4 flash can compete with established models like Claude Haiku 5.5 on satisfaction while offering significant cost savings. This could pressure other providers to lower prices or improve efficiency.

Read: GPT-6 Astra Automation

The availability of Hy4 flash on multiple platforms, including WorkBuddy, Yuanbao, and TokenHub, gives users flexibility in how they access the model. The planned release on OpenRouter will further expand its reach.

Tencent's move reflects a trend where AI companies are optimizing models for specific tasks rather than pursuing ever-larger models. This approach can lead to better performance and lower costs for practical applications.

The success of Hy4 flash will depend on how well it performs in real-world scenarios and whether users adopt it for their daily workflows. Tencent's promotional offers and weight release may accelerate adoption and provide valuable data for further improvements.

More from Inside AI

  • Features, Interviews, Press Releases

    Beyond Transcripts: Modulate Secures $25M to Scale Frontier Audio-Native AI Architecture Against Monolithic LLMs

    September 28, 2026
  • AI Policy & Regulation

    California Sues OpenAI Over Rogue AI Cyberattacks on Hugging Face

    November 17, 2026
  • Agentic AI

    Google Maps’ Ask Maps AI Can Now Plan Trips and Book Appointments

    November 16, 2026
  • AI Policy & Regulation

    Trump’s Nvidia Holdings Surge Nearly 100-Fold to $62 Million Amid Huang Scandal

    November 16, 2026
  • AI Policy & Regulation

    US Issues Rule to Protect Advanced AI Chips from Remote Access

    November 16, 2026
  • AI In Business

    Silicon Valley Leaders Back Petition to Reinstate Jensen Huang at Nvidia

    November 16, 2026
  • AI In Business

    OpenAI Brings Computer-Use Agents to ChatGPT for Healthcare

    November 16, 2026
  • AI In Business

    King’s Trust and OpenAI Launch £1m AI Skills Partnership for NEET Youth

    November 16, 2026
  • AI In Business

    Nvidia Earnings Test Interim CEO Kress After Huang Exit

    November 16, 2026

Never Miss a Breakthrough

Join 50,000+ readers who get our daily AI intelligence briefing. No fluff, just what matters.

Join Our Newsletter Community

Subscribe

Inside AI is an independent publication covering artificial intelligence news, machine learning research, and the tools shaping the future of technology. No hype. Just what's happening in the AI world.

Topics

  • Artificial Intelligence
  • Machine Learning
  • Generative AI
  • Agentic AI
  • Vibe Coding
  • Prompt Engineering
  • AI Policy & Regulation
  • AI Hardware & Infrastructure
  • AI Tools
  • AI In Business
  • Robotics
  • Cybersecurity AI
  • AI Safety
  • AI Tools & Reviews (Coming soon)

Company

  • Editorial Standards
  • Privacy Policy
  • Terms of Service
  • Contact
  • About Us

Others

  • Press Releases
  • Features
  • Sponsored Content
  • Advertise with us
  • Newsletter

© 2026 Inside AI. All rights reserved.

Designed by Blue Flare Digital