
一 | 持续了两年的AI硬件狂欢,终于进入了下半场。 2026年7月17日,工信部牵头的《人工智能终端智能化分级》系列国家标准首次公布测试结果,覆盖移动终端、微型计算机、电视、眼镜、汽车座舱、音箱六大品类,17家企业的66款产品通过L3辅助级认证。 手机品类的名单没有悬念,华为、小米、OPPO、vivo、荣耀、摩托罗拉、联想、智跃星辰,八大厂商共11款机型全部达到L3,主流手机品牌在AI能力上实现了全员入围,无人缺席。 The CAS held the first of two planned hearings of the Football Union of Russia and club teams’ appeal of a FIFA/UEFA ban from UEFA competition on Tuesday.,Tuesday’s hearing, taking place in Lausanne, Switzerland, will be followed by a second hearing next Monday by four Russian clubs, including Zenit St. Petersburg, the reigning Russian Premier League champions, challenging their exclusion from UEFA competition.,UEFA ruled in May to bar Russian football clubs from all tournaments, including men’s teams and the women’s team for the FIFA Women’s World Cup 2023. At least for the remainder of the current season, Russian teams have been banned from the UEFA Champions League, Europa League, Europa Conference League, Women’s Champions League, and Youth League. The Russian bid to host the Euro 2028 and/or Euro 2032 events has also been preemptively rejected.,In arbitration filed in March, the Football Union of Russia demanded the immediate reinstatement of all teams for competition, arguing the “irreparable harm” that their exclusion would have on tournaments, including the FIFA World Cup, potential sponsorship losses, loss of experience, and loss of opportunity to compete against opponents of high standard. The Russian side also stressed that for players 30 years old or older, the 2022 World Cup constituted “the last chance of their career to play such a tournament.”,Former Formula One Boss Ecclestone Backs Putin Over Conflict With Ukraine30 June, 12:39 GMT,The football union argued that the International Olympic Committee resolution issued in February demanding that Russian athletes and officials be barred from international competition on which FIFA and UEFA based their decisions was “generic,” “does not refer to any legal provision and is not binding upon FIFA,” and that sporting authorities were acting in a “discriminatory” manner by barring Russian players.,FIFA argued that Russian teams’ presence or absence in the World Cup and the Women’s World Cup was “not guaranteed,” and dismissed the fact that some players would miss out on their last chance for their World Cup play as “irrelevant.” Furthermore, the association argued that “pursuant to CAS jurisprudence, private sports associations have a wide autonomy and margin of discretion when deciding on matters related to the competitions which they organize.” The CAS sided with FIFA and UEFA and rejected Russia’s cases in April, prompting FUR to appeal.,US Hybrid War Against Russia,Rick Sterling, a California-based investigative journalist who has written extensively on doping in sports, said he finds it unlikely that the CAS will rule in Russia's favor despite its case's "excellent arguments," among them the idea that you don't punish athletes for the actions of their nations.,"The US invaded Iraq and there was no denunciation let alone a punishment of athletes. The US, France and others overthrew the Libyan government, leaving a destroyed state, without criticism from the IOC or FIFA. So why are the Russian athletes being punished? One reason is because of built-in bias and prejudice. The IOC and FIFA are dominated by the West. It is very damaging that the IOC issued a recommendation that Russian athletes be discriminated against. This is a violation of the Olympic Charter against national discrimination," Sterling says.,"The US is openly and actively working to undermine Russia's image abroad, including in sports. They probably pre-coordinated the response to the special operation. Russian athletes are victims of the US hybrid war against Russia, as are Ukrainians and Russians," the journalist believes.,Football isn’t the only sport to have suffered from politicized bans instituted by world and national sporting authorities in the wake of the escalation of the Ukraine crisis, with tennis, basketball, hockey, skating and skiing competitions also affected. The arbitrary nature of the prohibitions has been exemplified by their random nature, with Russian athletes barred from FIFA play, for example, but allowed to compete in the 2022 Winter Paralympics, and at the US Open tennis tournament.,Top Tennis Stars in Favour of Penalising Wimbledon Over Ban on Russian Players, Report Says11 May, 11:40 GMT,The bans affecting Russia’s football players are just the latest instance of discrimination against the country’s athletes by international sporting bodies in recent years. In 2020, the CAS stripped Russia of its right to use its name, flag, and anthem at the Olympics and world championship events for two years, largely on the basis of the testimony of Grigory Rodchenkov, the former director of the Russian Anti-Doping Center laboratory-turned-informant to the World Anti-Doping who fled to the United States and testified about an alleged Russian “state-run doping program.” Russian officials denied the allegations and accused Rodchenkov of altering doping probe data of Russian athletes to fit the required narrative.。

二 | 图源:《人工智能终端智能化分级》 除手机外,66款L3产品中,微型计算机独揽41款,电视9款,汽车座舱3款,眼镜和音箱各1款。部分眼镜、耳机仍停留在L2工具级。

三 | 从数量来看,AI终端智能化进程出现了明显的品类阶梯:PC领跑,手机紧随其后,穿戴设备仍在追赶。 雷科技(ID:leitech)认为,随着分级标准的落地,AI硬件狂欢正式进入下半场。 华米OV耀喜提L3,代表当前最高水平 先简单跟大家说说L1-L4的区别,以及为什么L5没有在分级名单中。

四 | L1 被称为响应级,终端仅具备被动响应能力,执行预设指令,如语音唤醒等,没有AI加持的Siri大概就是这个级别。L2 被称为工具级,终端可调用工具完成特定任务,初步具备场景感知能力,不过,在这个阶段用户指令仍需精确表述,模糊意图会导致执行失败,目前部分AI穿戴设备处于这一水平。 L3 是当前标准体系中有明确定义的最高等级,也叫辅助级。能实现复杂意图理解与主动澄清、任务自动分解与工作流编排、多工具动态调用与分布式任务协同、跨模态内容生成,以及短期和长期记忆能力。 图片来源:豆包 L4 是协同级,指向跨设备自主协同与主动场景感知,不过,现在对于L4的标准尚未完全明确,未来将随产业发展逐步完善。 值得注意的是,早期电信终端产业协会的《终端智能化分级研究报告》曾提出L5级别,即AI在感知和规划完成后,直接代为执行支付与下单操作,无需用户确认,但这一等级最终未被纳入国标。 原因在于主动服务的边界难以界定:比如当你对一款显卡犹豫不决时,AI直接帮你下单,那用户的数据安全与财务风险可能会失控。因此最终决策权应当保留在人类手中,原定的L5标准也可以看成是AI硬件的进化边界,可以辅助,但不能替代。 总而言之,目前L3代表的就是当前量产AI终端的真实技术水位,L4尚处于工程验证与标准讨论阶段。

五 | 从各家技术路线来看,华为搭载的小艺伴随式AI,强调端侧优先处理与隐私保护,能够在用户授权下主动提供服务而非被动等待唤醒;小米基于自研MiMo端侧大模型构建Xiaomi miclaw智能体,宣称可调用超过50项系统工具,支持跨应用、跨设备的多步骤任务编排;荣耀的YOYO智能体、vivo的蓝心大模型、OPPO的安第斯大模型,各自在端侧推理和多模态交互上投入了可观的研发资源。 图片来源:Android Central 但上述所有技术储备,目前的上限就是L3。这并非某一家厂商的问题,而是整个产业面临的技术性难题,端侧算力密度、模型推理效率、多Agent协同框架的成熟度,共同构成了L3到L4之间的技术壁垒。 巧的是,L4所指向的跨设备自主协同,恰好是各厂商发布会上反复描绘的未来图景,比如小米的人车家全生态、华为的1+8+N,框架早已搭建完成,但产品层面的实现尚需时日。 L3到L4两者之间的距离,就是AI终端产业未来三到五年的核心竞争区间。 PC拿下41款L3认证,白牌可能面临淘汰 让我们再回到《人工智能终端智能化分级》中,除手机外,其他品类也值得说道。 PC品类总共41款,相比其他有着压倒性的优势,主要原因是联想等厂商拥有完整的商用产品线和庞大的SKU矩阵,比如仅联想一家就贡献了37款AI PC,此外PC天然具备大算力芯片、充足散热空间和成熟的操作系统生态,是端侧大模型部署的最优载体之一。 科大讯飞副总裁娄超在WAIC现场也下了类似的判断:“PC一头连接国产算力、操作系统、大模型等技术创新,一头连接政务、教育、医疗和企业办公等实际需求,是推动AI终端产业发展的重要载体。” 图源:雷科技摄制 手机的挑战在于功耗比和散热,在手掌大小的空间内运行端侧大模型,对芯片能效比提出了极高要求。 同样受限于空间体积的还有穿戴设备品类,一副AI眼镜的目标重量通常控制在40至80克,在这一约束下能容纳的芯片算力、电池续航和散热能力都极其有限。目前行业主流的解决方案是端侧采集+云端/手机端处理的架构:眼镜负责语音收音、图像采集和交互入口,核心AI任务卸载至手机或云端完成。 但这不意味着AI眼镜的智能化进程滞后于市场需求。恰恰相反,AI眼镜是2026年增长最快的消费电子品类之一,IDC数据显示,2026年一季度全球出货量达356.6万台,同比增长130.1%。 用户对AI眼镜的接受度也比其他AI硬件高,据新华网报道,2026年春节期间,深圳华强北AI眼镜销量较平日激增70%至80%,经销商数量从零增至上百家,小雷此前去华强北也印证了这一点。 图片来源:千问 随着端侧AI芯片在能效比上的持续突破,以及MicroLED+光波导显示技术的成熟,眼镜品类从L2跨越到L3只是时间上的问题,节奏虽然有先后,但方向无差别。 可能AI分级对产业格局真正的冲击,是以华强北为代表的白牌AI硬件生态。 从家电圈能效等级的二十年历史来看,当分级标签成为消费者购买决策的核心依据,渠道体系会自发完成出清,电商平台和线下卖场没有动力让L2产品与L3产品同台竞争,白牌们可能要思考下一步该去向何方了。 标准画了一条线,但AI的想象力可以越界 当然,硬币也有另一面,L1到L4的分级框架下,相当于行业已经定义好了什么叫“智能”,你想达到L3,那就必须符合理解意图、拆解任务、多模态交互、持续记忆这些特征,不符合的就是L2甚至L1。 这套逻辑拿来衡量手机和PC,没问题,但AI硬件的想象力肯定远不止手机和PC。 回想一下过去两年那些不太像传统终端的AI产品,比如昙花一现的AI Pin,试图用激光投影取代屏幕,靠语音和手势完成所有交互;Rabbit R1用一个大模型操作APP的思路重新定义了什么叫用手机,虽然体验翻车,但这股探索的劲值得鼓励。 如果按现有的L1-L4标准来打分,这些产品大概率连L2都勉强,它们根本没有调用系统工具的框架,也不追求跨模态生成,它们探索的恰好是标准框架之外的可能性。 (图源:AI PIN/Humane) 不是小雷抬杠哈,标准天然是后验的,它总结的肯定是已经做出来的,但未来应该做什么,可能出现什么,都是未知数。 举个例子,2007年如果有人给智能手机下个标准,大概率会把全键盘、触控笔、物理导航键写进去,然后说没有这些不算智能。 但iPhone用一块玻璃屏幕把所有这些定义全推翻了,AI终端现在正处在类似的节点上,技术路线远未收敛,产品形态还在试错。

六 | 这时候用一套固定框架把所有产品装进四个格子里,装不进去的东西是不合格还是不在框架内?这两个说法之间的差距,就是品牌们可能会踩到的坑。

七 | 反过来说,没有标准也不行,过去两年“AI”标签被用滥了,从九块九的玩具到旗舰手机,什么东西都敢叫AI,分级至少让市场有了一个公认的标准,不至于人人AI。

八 | 如果未来L4、L5的标准修订能跟得上技术演进速度,新品类能及时被纳入分级框架,那它就是助推器。反过来,如果标准五年修订一次、品类列表十年不变,那它迟早变成一堵墙。 说到底,《人工智能终端智能化分级》开了一个好头,但能不能保持开放性,让下一批AI Pin和Rabbit R1在诞生时,不至于直接被贴上不合格的标签,这才是衡量这套标准格局的终极考题。 总而言之,这是中国AI硬件交出的第一份标准化答卷,但从目前来看,这份答卷的意义不在于分数本身,而在于它把“智能”从一个形容词变成了量化指标,从此以后,发布会上敢说自己AI领先的厂商,得先过标准这一关。 但大家也不必把标准当终点,PC和手机领跑L3,穿戴设备在追赶,品类之间的节奏差需要时间熨平,而标准框架本身的开放性,能不能接纳手机和PC之外的新物种,能不能在技术突变时快速迭代,决定了它是天花板还是起跑线。
Current article:http://30imu.zhouangnaoguangsundianzhenmou.cfd/w8e35t7/l3g.html
Published on:14:20:40