{"rewrite":{"id":"r_40e6083ef2b9e10568cbc24c","clusterId":"c_d12248e2e89ab255716e55c6","slug":"tencent-releases-smaller-open-model-hy3-matches-larger-rivals-in-scientific-tasks","model":"deepseek-v4-flash","headline":"Tencent Releases Smaller Open Model Hy3, Matches Larger Rivals in Scientific Tasks","summary":"Tencent released Hy3, an open-source AI model with 295 billion total parameters and 21 billion active parameters, on July 6. The model matches or outperforms larger models like GLM-5.2 and DeepSeek-V4-Pro in benchmarks, and beats GPT-5.5 in scientific research tasks. Hy3 is designed to reduce hallucinations by only answering when evidence exists. It is available under Apache 2.0, with API pricing from 1 yuan per million input tokens, and free on OpenRouter for two weeks.","whyItMatters":"Tencent's Hy3 demonstrates that a smaller open model can compete with much larger competitors in scientific reasoning, challenging the prevailing trend of scaling model size.","webCardHtml":"\u003cp\u003eTencent\u0026#39;s Hy3 follows a different design philosophy from many current large models. The developers implemented data cleaning and training constraints that instruct the model to only answer when it has evidence, and to state the lack of evidence when it does not. According to the company, this reduces confusion of facts, fabrication, and logical contradictions. The result is a 295-billion-parameter model that, in its internal document processing tests, used 47.4% fewer tokens than the larger GLM-5.2 model.\u003c/p\u003e\n\u003cp\u003eHy3 is released as an open model under the Apache License 2.0, and an API is available at 1 yuan per million input tokens. For a limited two-week period, access is free on OpenRouter.\u003c/p\u003e","blueskyPost":"Tencent's Hy3 open model matches larger rivals like GLM-5.2 and DeepSeek-V4-Pro in benchmarks, beats GPT-5.5 in scientific tasks. Uses evidence-based constraints to reduce hallucinations. 295B params, 21B active. Free on OpenRouter for 2 weeks.","twitterPost":"Tencent releases Hy3, an open-source AI model with 295B total params. Matches larger models in benchmarks, surpasses GPT-5.5 in scientific tasks. Reduces hallucinations by only answering when evidence exists. Free on OpenRouter for 2 weeks. API from 1 yuan/million input tokens.","threadsPost":null,"newsletterBlurb":"Tencent released Hy3, an open-source AI model with 295 billion total parameters and 21 billion active parameters, on July 6. The model matches or outperforms larger models like GLM-5.2 and DeepSeek-V4-Pro in benchmarks, and beats GPT-5.5 in scientific research tasks. Hy3 is designed to reduce hallucinations by only answering when evidence exists, and is available under Apache 2.0 with free access on OpenRouter for two weeks.","attributionJson":"[{\"source\":\"GIGAZINE\",\"url\":\"https://gigazine.net/news/20260707-tencent-ai-hy3/\",\"title\":\"Tencent releases AI model 'Hy3' as open model, 295B rivals GLM-5.2 and DeepSeek-V4, and surpasses GPT-5.5 in scientific tasks\"}]","lintFlagsJson":null,"lintHits":0,"costUsd":0,"inputTokens":4429,"outputTokens":2120,"status":"published","repairAttempts":0,"nextRepairAt":null,"factsAttemptedAt":1783391558,"createdAt":"2026-07-07T02:27:39.000Z","publishedAt":"2026-07-07T02:28:32.000Z","updatedAt":"2026-07-07T02:27:39.000Z"},"cluster":{"id":"c_d12248e2e89ab255716e55c6","canonicalTitle":"テンセントがAIモデル「Hy3」をオープンモデルとして公開、295BでGLM-5.2やDeepSeek-V4に匹敵＆科学タスクではGPT-5.5を上回る","representativeArticleId":"a_811175fcdbc39aa76a665d2d","sourceCount":1,"writtenSourceCount":1,"writeAttempts":0,"isSolo":true,"entitiesJson":"{\"anime_titles\":[],\"manga_titles\":[],\"work_titles\":[\"Hy3\"],\"studios\":[\"Tencent\"],\"people\":[],\"type\":\"announcement\",\"domain\":\"other\",\"is_roundup\":false}","contentType":"news","status":"published","firstSeenAt":"2026-07-07T01:51:00.000Z","lastSeenAt":"2026-07-07T01:51:00.000Z","updatedAt":"2026-07-07T02:28:32.000Z"},"attribution":[{"source":"GIGAZINE","url":"https://gigazine.net/news/20260707-tencent-ai-hy3/","title":"テンセントがAIモデル「Hy3」をオープンモデルとして公開、295BでGLM-5.2やDeepSeek-V4に匹敵＆科学タスクではGPT-5.5を上回る"}],"entities":{"anime_titles":[],"manga_titles":[],"work_titles":["Hy3"],"studios":["Tencent"],"people":[],"type":"announcement","domain":"other","is_roundup":false},"keyFacts":null}