Tencent's Hy3 AI model tops global rankings with 68x usage surge

Tencent's Hunyuan Hy3 large language model has opened to global developers, one week after its API call volume surged to 68 times that of its predecessor and claimed the top spot on OpenRouter's usage leaderboard.

The model processed 6.13 trillion tokens to rank No. 1 on OpenRouter's weekly leaderboard, according to the Tencent Hunyuan team. Hy3 launched simultaneously on open-source communities Hugging Face and ModelScope on its first day of release.

Tencent is offering free access to Hy3 through its WorkBuddy assistant until Aug. 31, 2026, targeting global developers, users and enterprises. The model has been integrated with platforms including Hermes, Kilo, Cline, OpenClaw, OpenCode and Cherry Studio.

The global rollout puts Tencent's cloud and AI business in direct competition with OpenAI, Anthropic and Google in the international LLM market. Tencent shares rose 1.3 percent on the Hong Kong exchange following the announcement.

Hy3's 6.13 Trillion Tokens Reshape the Leaderboard

The 68 times jump in API call volume from Hy2 to Hy3 reflects strong developer adoption, though the free-access window through WorkBuddy likely inflates usage metrics. OpenRouter's leaderboard measures raw token processing, which favors models offered at low or zero cost. Hy3's 6.13 trillion tokens processed in one week places it ahead of established models from OpenAI and Anthropic on the platform, though those companies distribute primarily through their own APIs rather than OpenRouter.

Tencent's decision to release Hy3 on Hugging Face and ModelScope on day one follows an open-source strategy similar to Meta's Llama releases, contrasting with the closed-weight approaches of OpenAI and Google. The integrations with developer tools like Cline, OpenCode and Cherry Studio suggest Tencent is targeting the coding and agentic AI segment, where competition with Anthropic's Claude and OpenAI's Codex is most intense. The model's rapid adoption also raises questions about whether the usage surge is sustainable once the free-access period ends, or whether developers will migrate to paid alternatives.

The global expansion of Hy3 also reflects a broader trend of Chinese AI companies pushing into international markets. Alibaba's Qwen models and DeepSeek's open-weight releases have already gained traction among global developers, and Tencent's entry intensifies the competitive pressure on US-based AI labs. The open-source distribution model gives Chinese AI companies a reach that closed-weight providers cannot easily match, particularly in price-sensitive emerging markets.

What the Free Access Window Means for Tencent's Cloud

Tencent's push into global AI comes as the company invests heavily in compute infrastructure. The free-access period through August 2026 will pressure near-term margins but could build long-term market share in the international LLM market. Tencent's cloud division stands to benefit if Hy3 converts usage into paid enterprise contracts, and the company has said Hy3 will be gradually integrated into more Tencent Cloud AI-native products.

Tencent shares rose 1.3 percent following the announcement, with short selling at HK$981.46 million representing an 8.1 percent ratio. BOCI maintained a cautious but flexible stance on Tencent despite the upsurge in AI investment, according to related coverage. The stock's positive reaction suggests investors view the global AI expansion as a growth driver for the company's cloud business, which competes with Alibaba Cloud and Huawei Cloud in the domestic market and with AWS and Microsoft Azure internationally. The key question for investors is whether Hy3's early momentum translates into durable revenue, or whether the free-access strategy simply buys temporary mindshare in a market where OpenAI and Anthropic hold entrenched enterprise relationships.

This article is for informational purposes only and does not constitute investment advice.