{"slug": "tencent-says-its-foundation-model-outperforms-rivals-in-internal-tests", "title": "Tencent says its foundation model outperforms rivals in internal tests", "summary": "Tencent claims its Hunyuan 3.0 foundation model, with 295 billion parameters and deep integration into WeChat and its Yuanbao AI assistant, outperforms rivals in internal tests, positioning it against Moonshot AI's Kimi K3, which launched July 17, 2026, with 2.8 trillion parameters and 1 million active tokens. Tencent is an investor in Moonshot AI, which is reportedly raising $2 billion at a $30 billion valuation, adding strategic tension to the domestic AI race.", "body_md": "Photo: Google DeepMind / Pexels\n\n# Tencent says its foundation model outperforms rivals in internal tests\n\nThe Chinese tech giant is positioning its Hunyuan series against one of its own portfolio companies in an increasingly heated domestic AI race.\n\nThe claim is particularly interesting because Tencent is also an investor in Moonshot AI. Funding your competitor and then trying to beat them is the kind of strategic tension that keeps corporate strategy consultants employed.\n\n## The players and their models\n\nTencent has been building out its Hunyuan model series, with the Hy3 (Hunyuan 3.0) release arriving in mid-2026. That model features 295 billion parameters and has been designed with a specific focus: deep integration into Tencent’s sprawling ecosystem of services, including WeChat and its AI assistant Yuanbao.\n\nOn the other side of this rivalry sits Moonshot AI, a company founded just in 2023. Moonshot launched its Kimi K3 model on July 17, 2026, and the specs are hard to ignore: 2.8 trillion parameters, support for 1 million active tokens via its mixture-of-experts (MoE) architecture, and competitive results against leading US models.\n\nMoonshot released open weights for Kimi K3 on July 27, 2026, roughly ten days after the initial launch.\n\n## The awkward investor dynamic\n\nMoonshot AI counts both Alibaba and Tencent among its major backers. The company is reportedly raising $2 billion at a valuation of approximately $30 billion, with plans for a potential Hong Kong stock listing.\n\nMoonshot’s rapid rise has drawn attention because its models have shown the ability to match or exceed some US counterparts on specific tasks, particularly in coding and web-related benchmarks. Independent evaluations have placed Kimi K3 ahead of or near some American AI models.\n\nAs of late August 2026, no explicit claims of outperformance between Tencent and Moonshot have been articulated in the media, indicating a cautious approach from both entities in public statements about direct competition.\n\n**Disclosure:** This article was edited by Editorial Team. For more information on how we create and review content, see our\n\n[Editorial Policy](https://cryptobriefing.com/editorial-policy/).", "url": "https://wpnews.pro/news/tencent-says-its-foundation-model-outperforms-rivals-in-internal-tests", "canonical_source": "https://cryptobriefing.com/tencent-foundation-model-moonshot-ai/", "published_at": "2026-08-28 11:14:25+00:00", "updated_at": "2026-08-28 11:21:44.436934+00:00", "lang": "en", "topics": ["artificial-intelligence", "large-language-models", "ai-products"], "entities": ["Tencent", "Hunyuan 3.0", "Moonshot AI", "Kimi K3", "WeChat", "Yuanbao", "Alibaba"], "alternates": {"html": "https://wpnews.pro/news/tencent-says-its-foundation-model-outperforms-rivals-in-internal-tests", "markdown": "https://wpnews.pro/news/tencent-says-its-foundation-model-outperforms-rivals-in-internal-tests.md", "text": "https://wpnews.pro/news/tencent-says-its-foundation-model-outperforms-rivals-in-internal-tests.txt", "jsonld": "https://wpnews.pro/news/tencent-says-its-foundation-model-outperforms-rivals-in-internal-tests.jsonld"}}