{"id":371745,"date":"2026-07-07T00:00:00","date_gmt":"2026-07-07T07:00:00","guid":{"rendered":"https:\/\/cms-articles.softonic.io\/en\/?p=371745"},"modified":"2026-07-07T00:00:35","modified_gmt":"2026-07-07T07:00:35","slug":"tencent-hy3-is-now-available-a-295b-open-moe-model","status":"publish","type":"post","link":"https:\/\/cms-articles.softonic.io\/en\/tencent-hy3-is-now-available-a-295b-open-moe-model\/","title":{"rendered":"Tencent Hy3 is now available: a 295B open MoE model"},"content":{"rendered":"<p class=\"wp-block-paragraph\">Tencent released Hy3 on July 6, 2026. It\u2019s a new open Mixture-of-Experts model with <strong>295B total parameters<\/strong>, 21B active on any given task, a 256K context window, and an Apache 2.0 license.<\/p>\n\n<p class=\"wp-block-paragraph\">The MoE design is the point here: each prompt only passes through part of the Hy3 network, which Tencent says cuts <strong>inference and serving costs<\/strong>. Hy3 is also meant to stay useful on long documents, large codebases, and multi-step work. And because it\u2019s under Apache 2.0, you\u2019ve got wide commercial freedom to self-host it, fine-tune it, and build global products with it, without the geographic restrictions that showed up in some earlier China-based open models.<\/p>\n\n<p class=\"wp-block-paragraph\">Tencent says the July 6, 2026 release is <strong>better than the April preview<\/strong> after feedback from 50 internal product teams. By Tencent\u2019s numbers, hallucinations dropped from 12.5% to 5.4%, and commonsense errors fell from 25.4% to 12.7%. Hy3 is already being used in WorkBuddy, <a href=\"https:\/\/yuanbao.en.softonic.com\/\" rel=\"noopener\">Yuanbao<\/a>, <a href=\"https:\/\/wechat.en.softonic.com\/\" rel=\"noopener\">WeChat assistants<\/a>, and Path of Exile: Advent, which makes it feel more substantial than a research demo built mainly to post benchmark scores.<\/p>\n\n<p class=\"wp-block-paragraph\">If you want an open model for search, tool use, or agentic tasks, Hy3 looks like one to download. Coding may still be a weaker spot, since Zhipu AI\u2019s GLM-5.2 could still have the edge there, and Tencent\u2019s claims need <strong>independent testing<\/strong> either way. Still, Hy3 now joins Zhipu AI, <a href=\"https:\/\/deepseek-tui.en.softonic.com\/\" rel=\"noopener\">DeepSeek<\/a>, Alibaba\u2019s Qwen, and Mistral AI in putting more pressure on closed models from <a href=\"https:\/\/en.softonic.com\/articles\/openai-and-google-get-ready-microsoft-is-launching-its-most-powerful-ai-models-to-date\" rel=\"noopener\">OpenAI and Anthropic<\/a>.<\/p>\n\n<p class=\"wp-block-paragraph\">You can get Hy3 from Tencent as an open release under the Apache 2.0 license.<\/p>","protected":false},"excerpt":{"rendered":"<p>Tencent released Hy3 on July 6, 2026. It\u2019s a new open Mixture-of-Experts model with 295B total parameters, 21B active on any given task, a 256K context window, and an Apache 2.0 license. The MoE design is the point here: each prompt only passes through part of the Hy3 network, which Tencent says cuts inference and &hellip; <a href=\"https:\/\/cms-articles.softonic.io\/en\/tencent-hy3-is-now-available-a-295b-open-moe-model\/\" class=\"more-link\">Continue reading<span class=\"screen-reader-text\"> &#8220;Tencent Hy3 is now available: a 295B open MoE model&#8221;<\/span><\/a><\/p>\n","protected":false},"author":9265,"featured_media":371744,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":"","wpcf-pageviews":0},"categories":[1015],"tags":[],"usertag":[],"vertical":[],"content-category":[6771],"class_list":["post-371745","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-news","content-category-ai"],"aioseo_notices":[],"_links":{"self":[{"href":"https:\/\/cms-articles.softonic.io\/en\/wp-json\/wp\/v2\/posts\/371745","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/cms-articles.softonic.io\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/cms-articles.softonic.io\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/cms-articles.softonic.io\/en\/wp-json\/wp\/v2\/users\/9265"}],"replies":[{"embeddable":true,"href":"https:\/\/cms-articles.softonic.io\/en\/wp-json\/wp\/v2\/comments?post=371745"}],"version-history":[{"count":1,"href":"https:\/\/cms-articles.softonic.io\/en\/wp-json\/wp\/v2\/posts\/371745\/revisions"}],"predecessor-version":[{"id":371746,"href":"https:\/\/cms-articles.softonic.io\/en\/wp-json\/wp\/v2\/posts\/371745\/revisions\/371746"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/cms-articles.softonic.io\/en\/wp-json\/wp\/v2\/media\/371744"}],"wp:attachment":[{"href":"https:\/\/cms-articles.softonic.io\/en\/wp-json\/wp\/v2\/media?parent=371745"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/cms-articles.softonic.io\/en\/wp-json\/wp\/v2\/categories?post=371745"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/cms-articles.softonic.io\/en\/wp-json\/wp\/v2\/tags?post=371745"},{"taxonomy":"usertag","embeddable":true,"href":"https:\/\/cms-articles.softonic.io\/en\/wp-json\/wp\/v2\/usertag?post=371745"},{"taxonomy":"vertical","embeddable":true,"href":"https:\/\/cms-articles.softonic.io\/en\/wp-json\/wp\/v2\/vertical?post=371745"},{"taxonomy":"content-category","embeddable":true,"href":"https:\/\/cms-articles.softonic.io\/en\/wp-json\/wp\/v2\/content-category?post=371745"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}