[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-4-china-labs-12-days-coding-open":3,"news-related-88d40bcd-f92f-481e-b644-5da3dd9813a4":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},"88d40bcd-f92f-481e-b644-5da3dd9813a4","四家中国实验室十二天内密集发布开源代码模型：前沿能力与低成本并行","2026年4月下旬，智谱AI、MiniMax、Moonshot AI与DeepSeek四家中国实验室在短短12天内相继发布开源代码模型，密度创开源模型历史纪录。这一连串发布不仅是参数规模的堆砌，更在编程与推理能力上展现出真正的前沿竞争力。\n\nKimi K2.6发布于4月24日，采用万亿参数MoE架构。发布后不久，K2.6在AI编程挑战赛中超越GPT-5.5、Claude Opus 4.7与Gemini系列，拿下第一名，引发行业震动。这一结果打破了「开源模型性能必然落后于闭源前沿」的固有认知，证明通过长时推理优化与稀疏注意力机制，小型开源模型也能在特定任务上与最强闭源模型正面竞争。\n\nMiniMax于4月22日推出M2.7，采用「自进化」训练路径——模型参与自身的训练过程，通过持续反思与优化迭代提升能力。这种方法跳出了传统的固定预训练范式，为模型训练开辟了新思路。\n\n随着模型能力趋同，推理效率正在成为新的竞争维度。DeepSeek V4通过稀疏注意力机制，将长上下文推理成本压缩数倍；Kimi K2.6借助MoE架构，在保持万亿参数规模的同时控制推理计算量。2026年的开源模型战场，正在从「谁参数大」转向「谁架构更聪明」。","https:\u002F\u002Fwhatllm.org\u002Fblog\u002Fnew-ai-models-may-2026","cae10e40-cce7-44e5-91c1-fa1699026237",[10,14,17,20],{"id":11,"name":12,"slug":12,"description":13,"color":13},"a8002d98-9df1-4ab9-94d4-a7625af634c4","china-ai",null,{"id":15,"name":16,"slug":16,"description":13,"color":13},"01598627-1ea6-4b27-a5d8-874971571a71","llm",{"id":18,"name":19,"slug":19,"description":13,"color":13},"b1853a5a-d940-42b7-94f9-0488ee3f2cf7","new-model",{"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},"8caedd5a-17e8-49fc-8c66-687c56696bc1","en","Four Chinese labs ship open coding models in twelve days","In late April 2026, four Chinese labs — Zhipu AI, MiniMax, Moonshot AI, and DeepSeek — released open-source coding models one after another in just 12 days, a density that set an open-source-model historical record. This string of releases isn't just parameter-count piling up, but a real display of frontier-level competitiveness in coding and reasoning.\n\nKimi K2.6 was released on April 24, using a trillion-parameter MoE architecture. Shortly after launch, K2.6 surpassed GPT-5.5, Claude Opus 4.7, and the Gemini family in an AI programming challenge, taking first place and sending shockwaves through the industry. This result shatters the long-held belief that \"open-source models must lag behind closed-source frontiers,\" proving that through long-horizon reasoning optimization and sparse attention mechanisms, small open-source models can also go head-to-head with the strongest closed-source models on specific tasks.\n\nMiniMax launched M2.7 on April 22, adopting a \"self-evolution\" training path — the model participates in its own training process, with continuous reflection and iterative optimization improving capability. This approach breaks out of the traditional fixed-pretraining paradigm, opening new directions for model training.\n\nAs model capabilities converge, inference efficiency is becoming the new competitive axis. DeepSeek V4 uses sparse attention to compress long-context inference cost several-fold; Kimi K2.6 leverages MoE architecture to control inference compute while maintaining trillion-parameter scale. The 2026 open-source model battlefield is shifting from \"who has more parameters\" to \"who has the smarter architecture.\"","4-china-labs-12-days-coding-open","2026-05-21T02:30:00Z","2026-05-21T10:12:15.291307Z","2026-08-19T02:08:40.142862Z",true,"agent",101,{"items":37},[38,43,48,53,58,63],{"id":39,"title":40,"news_slug":41,"published_at":42},"010979e2-4e0a-4dcc-8a91-c6a99bfebc04","腾讯混元 Hy3 Preview 开源：295B MoE 剑指 Agent 实用性","tencent-hunyuan-hy3-preview-295b-moe","2026-05-08T13:00:00+00:00",{"id":44,"title":45,"news_slug":46,"published_at":47},"6836f7c8-c430-4722-afaa-35d95f40e100","开源LLM的崛起：从追赶引领到标准制定","open-source-llm-rising-china-qwen-glm-deepseek","2026-04-25T02:05:00+00:00",{"id":49,"title":50,"news_slug":51,"published_at":52},"fc0cd2d4-cc5f-4f97-b91f-08719e41e8ec","Qwen3.6-27B：27B密集模型超越397B MoE，单卡部署的编程新选择","qwen-3-6-27b-dense-beats-397b-moe-coding-77pct","2026-04-24T03:30:00+00:00",{"id":54,"title":55,"news_slug":56,"published_at":57},"d4d40e4b-04e9-45cf-a04c-792aca45b152","Kimi K2.6开源发布：万亿参数MoE模型的长时编程与Agent Swarm突破","kimi-k2-6-trillion-moe-1t-32b-active-256k","2026-04-24T03:00:00+00:00",{"id":59,"title":60,"news_slug":61,"published_at":62},"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":64,"title":65,"news_slug":66,"published_at":67},"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"]