[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-tencent-hunyuan-hy3-openrouter":3,"news-related-dfa4fd62-754c-4c08-b7c4-5f8bc4c7f678":36},{"id":4,"title":5,"summary":6,"content":6,"original_url":7,"source_id":8,"tags":9,"translations":23,"news_slug":29,"published_at":30,"created_at":31,"modified_at":32,"is_published":33,"publish_type":34,"image_url":13,"view_count":35},"dfa4fd62-754c-4c08-b7c4-5f8bc4c7f678","上线一周调用量增 68 倍,腾讯混元 Hy3 在 OpenRouter 全球登顶","腾讯混元大模型 Hy3 自 7 月初正式开源以来表现远超预期——7 月 15 日官方披露,Hy3 在 OpenRouter 全球大模型调用量总榜登顶,总调用量较上代 Hy2 暴涨 68 倍。\n\nHy3 是 295B 总参数、激活仅 21B 的稀疏 MoE 模型,GitHub 自定位 \"reasoning and agent\",主打同尺寸段的推理与 Agent 能力,同时强调成本效率。A21B 这种高稀疏度设计意味着推理时实际激活参数不到总量的 7%,在保留能力的同时把单次调用成本压到\"接近稠密 20B 级\"。这也是它在 OpenRouter 这种\"性价比投票\"聚合平台上能跑出来的根本原因——调用方用脚投票,价格\u002F能力比才是关键。\n\n68 倍增长与 OpenRouter 登顶本质上是开发者社区对\"中国系 MoE + Agent\"组合的一次真实压力测试,而非纸面 benchmark。头部大厂已把\"调用量\"作为开源成功的新指标,意味着 LLM 竞争正从\"谁能训出来\"转向\"谁敢被广泛调用\"。对中型闭源 API 提供商而言,Hy3 这类高性价比开源 MoE 已构成直接挤压。","https:\u002F\u002F36kr.com\u002Fnewsflashes\u002F3896804710647431","d46ec0a7-501b-4ef8-9c89-2391b2701b3b",[10,14,17,20],{"id":11,"name":12,"slug":12,"description":13,"color":13},"120fa59a-ff6f-4537-9bf5-f818df636a0e","benchmark",null,{"id":15,"name":16,"slug":16,"description":13,"color":13},"a8002d98-9df1-4ab9-94d4-a7625af634c4","china-ai",{"id":18,"name":19,"slug":19,"description":13,"color":13},"01598627-1ea6-4b27-a5d8-874971571a71","llm",{"id":21,"name":22,"slug":22,"description":13,"color":13},"b9bd9039-fcdb-41a8-b85b-fc1587def2b9","open-source",[24],{"id":25,"lang":26,"title":27,"summary":28,"content":13},"7bf0d41e-f1ae-4cec-babd-985b5198eab6","en","Tencent Hy3 tops OpenRouter globally, 68x calls in one week","Tencent's Hunyuan large model Hy3 has performed far beyond expectations since its open-source release in early July — on July 15, the company disclosed that Hy3 has topped OpenRouter's global LLM call-volume leaderboard, with total call volume up 68× over its predecessor Hy2. Hy3 is a 295B-total, 21B-activated sparse MoE model, with GitHub self-positioning as \"reasoning and agent\", focused on reasoning and Agent capabilities in the same size tier, while emphasizing cost efficiency. This A21B high-sparsity design means that at inference, the actually activated parameters are less than 7% of the total, preserving capability while compressing per-call cost to \"close to a dense 20B tier\". This is fundamentally why it can emerge on OpenRouter, a \"price-performance vote\" aggregation platform — callers vote with their feet, and the price\u002Fcapability ratio is what matters. The 68× growth and OpenRouter topping are essentially a real-world stress test of the \"Chinese MoE + Agent\" combination by the developer community, rather than paper benchmarks. The leading labs have made \"call volume\" a new metric for open-source success, meaning LLM competition is shifting from \"who can train it\" to \"who dares be widely called\". For mid-sized closed-source API providers, high-price-performance open-source MoE models like Hy3 already constitute direct pressure.","tencent-hunyuan-hy3-openrouter","2026-07-16T00:01:00Z","2026-07-16T00:05:50.290972Z","2026-08-19T02:08:40.142862Z",true,"agent",207,{"items":37},[38,43,48,53,58,63],{"id":39,"title":40,"news_slug":41,"published_at":42},"ad3be632-49a1-44f5-9816-62c168e56467","全球大模型调用量榜前五全是\"中国造\":开源 MoE 正在重写 OpenRouter 的地理坐标","openrouter-top5-china-moe-open-source-2026w31","2026-08-02T03:30:00+00:00",{"id":44,"title":45,"news_slug":46,"published_at":47},"37c8dc68-276b-4905-baed-1e120e86287c","小米MiMo-V2.5登顶OpenRouter周榜月榜：国产开源MoE拿下全球调用量第一","mimo-v2-5-openrouter","2026-07-28T03:00:00+00:00",{"id":49,"title":50,"news_slug":51,"published_at":52},"f6e4aab0-7693-4c2c-bb66-c1641fc2cc3e","Ox Alpha 谜底揭晓:智谱 GLM-5.3-Flash,MIT 开源 320B MoE","ox-alpha-glm-5-3-flash-reveal","2026-08-27T13:30:00+00:00",{"id":54,"title":55,"news_slug":56,"published_at":57},"804ab59a-a8d6-4b61-bf74-8f6f2bdae83c","智谱把 Flash 做成一件正经事:一次说清 GLM-5.3-Flash 的架构和 benchmark 真相","glm-5-3-flash-hybrid-attention-architecture","2026-08-27T08:00:00+00:00",{"id":59,"title":60,"news_slug":61,"published_at":62},"68072ee1-fc37-4064-ab18-09550ae72d1b","GLM-5.3-Flash 把 320B MoE 跑在国产芯片上:Flash 价位和 $0.15 API 的混合注意力栈","glm-5-3-flash-chinese-chips-hybrid-attention","2026-08-27T03:00:00+00:00",{"id":64,"title":65,"news_slug":66,"published_at":67},"ff0bc92a-295a-4707-be8d-76115fe9eeee","PerceptionBench 出炉:16 个前沿多模态模型,视觉感知无一及格","moonshot-perceptionbench-atomic-perception","2026-08-26T13:15:00+00:00"]