[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-673407a6-0253-4add-839d-845f6baad077":3},{"id":4,"title":5,"summary":6,"original_url":7,"source_id":8,"tags":9,"published_at":23,"created_at":24,"modified_at":25,"is_published":26,"publish_type":27,"image_url":13,"view_count":28},"673407a6-0253-4add-839d-845f6baad077","Gemini 3.5 Pro推迟发布：Google I\u002FO 2026的两点观察","Gemini 3.5 Pro推迟发布：Google I\u002FO 2026的两点观察\n\nGoogle I\u002FO 2026上，Gemini 3.5 Flash如约而至，而Gemini 3.5 Pro却被推迟到6月。Sundar Pichai那句\"Give us until next month\"换来现场一片叹息。但仔细看基准测试数据，这次推迟并非能力不足，而是有明确的技术逻辑。\n\n**Flash的强项与软肋**\n\nGemini 3.5 Flash在编程和Agent基准上大幅超越前代3.1 Pro：Terminal-Bench提升5.9%、MCP Atlas提升5.4%、Finance Agent v2提升14.9%。Google选择把Flash定位成默认模型，正是因为它在中轻度任务上已经足够强——速度是前代4倍，成本下降40%。\n\n但Flash也暴露了三项退步：Humanity's Last Exam下降4.2%、ARC-AGI-2下降5.0%、128K长上下文检索下降7.6%。这三项对应深度推理和长文档处理，是当前竞争最激烈的能力高地。\n\n**Pro的任务：精准补位**\n\nGemini 3.5 Pro的发布，本质上是对这次退步的精准补救。如果Pro能恢复3.1 Pro在这三项上的表现，同时保留Flash的编程能力，它将成为目前最均衡的旗舰模型——既有速度，又有深度。\n\n定价悬念仍在——若高于3.1 Pro水平，可能强化其高端推理定位。这次推迟折射出Google的战略转向：从追求单点能力峰值，转向在速度、成本与能力之间寻找最优平衡点。","https:\u002F\u002Fwavespeed.ai\u002Fblog\u002Fposts\u002Fgemini-3-5-pro-coming-next-month\u002F","90791fcd-1d9f-4f06-a676-0673fd491bce",[10,14,17,20],{"id":11,"name":12,"slug":12,"description":13,"color":13},"a9524a82-a7c5-4daa-bb4b-a7ee77bb0b94","gemini",null,{"id":15,"name":16,"slug":16,"description":13,"color":13},"8cf7490f-2449-4ba7-be19-61befa0d92b4","google",{"id":18,"name":19,"slug":19,"description":13,"color":13},"0a93ec8e-ea39-4693-81de-563ca8c173f7","inference",{"id":21,"name":22,"slug":22,"description":13,"color":13},"01598627-1ea6-4b27-a5d8-874971571a71","llm","2026-06-01T08:15:00Z","2026-06-01T16:08:43.114931Z","2026-06-01T16:08:43.114944Z",true,"agent",3]