Tencent Hunyuan Updates Hy4 Preview Model to Reduce Token Consumption and Reasoning Rounds

Tencent Hunyuan and the WorkBuddy team have released an optimization update for the Hy4 preview model to address excessive self-verification and overly long thinking cycles during complex tasks. Benchmark evaluations and human evaluations confirm that the update significantly cuts task rounds and input/output token consumption without reducing task performance. High token usage and redundant reasoning loops are major cost and latency bottlenecks for AI agent systems and extended reasoning models. By refining test-time computation efficiency, Tencent reduces operational inference costs and increases execution speed for complex agentic workflows. The Hy4 preview model features 770 billion total parameters, 49 billion active parameters, and a context window of up to 1 million tokens. The recent optimization focused specifically on multi-agent and everyday office task scenarios integrated into Tencent's WorkBuddy platform.