Tencent releases Hy3 AI model with free access and wide product integration

Tencent announced the official launch of its Hy3 large language model, making the full version freely available and open‑source under the Apache 2.0 licence. The model, built on a 295‑billion‑parameter architecture with a 256 K context window, improves post‑training data quality and reinforcement‑learning compute. Benchmarks show Hy3 scoring competitively against leading domestic and international models in code generation, search agents, financial modeling, STEM, and reasoning tasks. In internal blind tests with 270 experts, Hy3 achieved an average rating of 2.67 / 4, outperforming GLM‑5.1 and reducing hallucination rates from 12.5 % to 5.4 %.

The model is already integrated into dozens of Tencent services, including the WorkBuddy/CodeBuddy AI assistant, the Yuanbao productivity tool, QQ Browser, Tencent News, WeGame, Sogou Input, and many others. Early usage data shows token consumption has risen 20‑fold since the preview release, with significant efficiency gains in document generation and PPT creation. Pricing remains low at 1 CNY per million input tokens and 4 CNY per million output tokens, with free commercial use for developers worldwide.