Home
Tencent Unveils Open-Source Hunyuan Hy4preview, Flagship Model Boasting 770B Parameters and 1M Context Window
Tencent Hunyuan launched its new open-source flagship large model, Hy4preview, on August 28. Featuring 770B total parameters and 49B activated parameters, the model supports a 1M context window and is designed for real-world productivity applications. It has been open-sourced across HuggingFace, GitHub, and Modelscope, and is accessible via Tencent Cloud TokenHub and OpenRouter. Users can test the model directly in tools like WorkBuddy, CodeBuddy, Yuanbao, and ima.

Hy4preview demonstrates significant advancements in four key areas, leveraging high-quality data curated with experts from Tencent’s software engineering, gaming, finance, and security divisions. These improvements include enhanced understanding, planning, and debugging for long-term software development tasks; end-to-end delivery capabilities in office scenarios, spanning data analysis to documents, spreadsheets, and presentations; support for Unreal5 and Unity engines via MCP in game development, enabling the generation of playable demos through natural conversation; and breakthroughs in scientific research, including molecular dynamics, condensed matter physics, and advanced mathematics.

Internal blind tests involving 203 engineering tasks evaluated by 163 experts show Hy4preview achieving an average score of 2.99/4.00, outperforming GLM-5.3 (2.92) and Kimi K3 (2.94). In scientific research, paired with the Hyra model, it improved the volume lower bound for the three-dimensional Blaschke–Lebesgue problem from 0.380799 to 0.41104, narrowing the gap to the Meissner tetrahedron conjecture proof to just 2%. Additionally, it achieved a 2.0x speedup in simulating a 32,512-atom phospholipid bilayer, reaching 54.9ms per step.
Tencent describes Hy4preview as an early iteration of the Hy4 series. While pre-training and post-training offer room for further optimization, known limitations such as extended reasoning for complex tasks remain. The model will evolve through agile iterations based on real user feedback. This open-source early release strategy reflects a broader industry shift from competing on parameter counts to validating practical productivity effectiveness.
Related article
U.S. Stocks Hit Historic Milestone as AI and Aerospace Giants Prepare for Trillion-Dollar Debut
Elon Musk, Sam Altman, and Dario Amodei, three titans of the technology sector, are advancing toward initial public offerings for their respective ventures. With SpaceX, OpenAI, and Anthropic—three industry behemoths nearing trillion-dollar valuation
Swedish AI Startup Lovable Eyes $13.2 Billion Valuation After Major Funding Round
As AI-driven coding tools gain traction, Swedish startup Lovable has secured a major funding round. The company aims to raise $3 billion, potentially boosting its valuation to $13.2 billion—double the $6.6 billion recorded last December. Menlo Ventur
Google Tests Remy AI Agent for Gemini as Focus Shifts to User Control
According to Business Insider, Google is testing Remy, a new AI personal agent for Gemini. This tool aims to execute tasks on behalf of users, streamlining both professional workflows and daily routines.Currently, Remy is undergoing testing in an int
Related Special Topic Recommendations
Comments (0)
0/500
Tencent Hunyuan launched its new open-source flagship large model, Hy4preview, on August 28. Featuring 770B total parameters and 49B activated parameters, the model supports a 1M context window and is designed for real-world productivity applications. It has been open-sourced across HuggingFace, GitHub, and Modelscope, and is accessible via Tencent Cloud TokenHub and OpenRouter. Users can test the model directly in tools like WorkBuddy, CodeBuddy, Yuanbao, and ima.

Hy4preview demonstrates significant advancements in four key areas, leveraging high-quality data curated with experts from Tencent’s software engineering, gaming, finance, and security divisions. These improvements include enhanced understanding, planning, and debugging for long-term software development tasks; end-to-end delivery capabilities in office scenarios, spanning data analysis to documents, spreadsheets, and presentations; support for Unreal5 and Unity engines via MCP in game development, enabling the generation of playable demos through natural conversation; and breakthroughs in scientific research, including molecular dynamics, condensed matter physics, and advanced mathematics.

Internal blind tests involving 203 engineering tasks evaluated by 163 experts show Hy4preview achieving an average score of 2.99/4.00, outperforming GLM-5.3 (2.92) and Kimi K3 (2.94). In scientific research, paired with the Hyra model, it improved the volume lower bound for the three-dimensional Blaschke–Lebesgue problem from 0.380799 to 0.41104, narrowing the gap to the Meissner tetrahedron conjecture proof to just 2%. Additionally, it achieved a 2.0x speedup in simulating a 32,512-atom phospholipid bilayer, reaching 54.9ms per step.
Tencent describes Hy4preview as an early iteration of the Hy4 series. While pre-training and post-training offer room for further optimization, known limitations such as extended reasoning for complex tasks remain. The model will evolve through agile iterations based on real user feedback. This open-source early release strategy reflects a broader industry shift from competing on parameter counts to validating practical productivity effectiveness.
U.S. Stocks Hit Historic Milestone as AI and Aerospace Giants Prepare for Trillion-Dollar Debut
Elon Musk, Sam Altman, and Dario Amodei, three titans of the technology sector, are advancing toward initial public offerings for their respective ventures. With SpaceX, OpenAI, and Anthropic—three industry behemoths nearing trillion-dollar valuation
Swedish AI Startup Lovable Eyes $13.2 Billion Valuation After Major Funding Round
As AI-driven coding tools gain traction, Swedish startup Lovable has secured a major funding round. The company aims to raise $3 billion, potentially boosting its valuation to $13.2 billion—double the $6.6 billion recorded last December. Menlo Ventur











