Summary
Tencent open-sourced Hy4 Preview, a 770B parameter mixture-of-experts model with about 49B active parameters and 1 million token context. It excels on coding benchmarks: Terminal Bench 2.1 score 85.4 (tied with Claude Opus 5, surpassing DeepSeek V4 Pro), DeepSWE 64.3, SWE-Bench Pro 65.7, MCP-Atlas…
Key points
- Provides developers with an open-source, long-context, high-performance coding model choice, understanding its pros and cons to decide whether to adopt.
- Tencent's Hy4 is a significant advance in Chinese open-source models, offering an alternative to closed-source models, but actual performance needs verification.
- Developers can leverage Hy4 for long-context coding tasks but must tolerate latency; its open-source nature may reduce reliance on Western models, impacting China's AI ecosystem.
Editorial note
This page is Code & Chain's editorial summary of public sources. It may be prepared with AI assistance and published through an automated workflow. Refer to the original sources; this content is not investment, legal, or tax advice.