{ "text": "各位朋友大家好\n欢迎来关注昆蓬道今天告诉大家一个好消息\nDeepseek v4.1 flash 这个模型正式开源了\n那他目前已经把他的模型权重上传到huginface了\n然后官方提供的是一个\n应该是一个FP8的一个格式的一个模型\n模型权重的510个G\n我们来看一下他相关的一些参数\n那本次他是一个552B的一个大模型\n也是一个MOE的架构\n然后他本次调整的那个是他的一个CDE的架构\n就是由原来的40层的转换器\n然后现在优化成了改成20层的英国编码器盒\n后面接了20层的解码器\n这样的话能够实现前局的KV缓存来进行加速\n就是让它能够实现就是预填充的时候\n它是一个激活参数是8B的\n然后就是解码的时候它的速度是16B的\n这样的话又实现了一个全新的一个技术架构\n能够实现大模型的一个推理速度加速\n然后本次他的这一次的模型\n已经全量在他们官网上线了\n而且他官网已经把Deepseek v4 Pro\n已经路由到他的Deepseek v4 Flash 4.1的模型上面来了\n而且他经过他的架构优化\n能够让他的kv缓存降低了\n就是降低到原来的1%\n另外一个就是他用了一个压缩的吸收注意力的算法\n然后能够实现也是一个主要是用来加速的\n在各个层级用的吸收索引器\n然后来进行限制索引\n然后实现的一个推理的一个加速\n我觉得这块确实在DeepSick在这方面\n还是做了很多的一个功夫\n另外的一个就是它的一个MHC\n这个之前他已经在他的开源的项目里面\n已经多次提到过\n包括他的Ingram的一个存储器\n然后很早的时候应该是在\n25年11月份左右的时候\n它开源出来的\n另外一个就是它本次是一个全新的\n完整的多模态的一个大模型\n它跟原来Deepseek v4 flash\n相当于在前面加了一个补丁包\n不一样的是它本次的多模态\n是用了Deepseek v8的一种技术架构\n然后它是从预训练的时候\n就已经开始用到了多模态的一个语料\n来进行训练\n所以它说它是一个原生的多模态大模型\n这个是没有问题的\n那我们可以看一下它的一个分数\n它本次也很克制\n然后它在TerminalBrain Center上面的话\n能拿到了一个30分\n然后比Kimi5.3都要好一些\n然后比Opuls5还是差的很多\n我们可以看一下\n然后和GPT5.6\n那其实GPT6一出来了\n所以说还是有很大的一个差距\n在国内在TerminalBrain上面\n当然这些这叫什么\n这些压测的测试的数据集\n包括算法都是国外的他们做的\n有可能都是由open ai或者说cloud\n他们定义的\n就谁定义谁说算\n他肯定的数据会好一些\n当然了这个不能绝对\n人家的模型确实我们承认他比较先进\n另外一个在这个Deep SWE上面\nV1.1的这个版本测试数据以上的话\n他应该是拿到了一个最好的一个成绩\n然后在安全的这个测评的上面\n也拿到一个88分\n然后在自动化上面的也能拿到一个54分\n目前我看到网上很多的用户评论说\n把它接到Depseek里面去目前\n不是\n接到Codex里面去用Codex Computer Use\n是目前性价比和成功率最高的模型之一\n除了Depseek\n除了GPT官网之外\n然后本次算是Depseek真正的机身\n然后世界前沿级的大模型\n多模态加一照上下文的这个梯堆\n目前国内相当于是第三家\n第一家相当于是Kimi\n第二家相当于是千问\n但千问38刚开始推出来的不是多么泰的\n我也很奇怪\n不知道他现在是不是\n我们就把它放在这一列里面\n因为我之前说国内只有一家的时候\n就有人在评论区说千问也是\n我记得千问38刚出来的时候\n它是一个文本的多么泰\n因为大家都知道\n我其实没有做测试\n没有深入的测试\n然后限制太多了\n所以我就放弃了\n然后这个是它的整个的一个测评的一个分数\n比如说在各个领域\n然后还是非常亮眼的\n这个黑色的就表示\n它是目前的在已知的测评的这些里面\n效果是\n效果是最好的\n那我们可以看一下\n这个是它的一个base模型和它的\n对就是这个模型\n它在很多的参数已经非常接近v4Pro\n和已经超过的Pro的模型\n所以说他们官网已经把这个\n已经路由到Pro路由到这个模型上面来\n第一个是为了降低它的一个推理成本\n第二个也是因为这个模型的能力确实提升非常明显\n还有一个的最主要原因是网上被很多用户吐槽\nDeepstick v4Pro的幻觉比较严重\n实际上我使用下来整体感觉还是非常不错的\n他这些测评的分数我们就不一一讲了\n因为刚才我们已经带大家看了一下\n他跟国外的这些模型的做一些对比\n其实我们可以看到目前他应该是\n激升到了国内的大模型的地体对\n毫无疑问\n所以说DeepSick还是非常强的\n目前他官方还没有去推出\n然后VM和SGLAN部署的\n然后我相信应该很快\nVMHSG浪就有对您的支持\n就算没有对您也有对E的支持\n刚才我们在一个我们私有化部署大木型的群里\n就有人说今天晚上想蹲一个40E的本地部署\n我感觉够呛\n其实我们手上也有机器\n但是不打算部署了\n因为它500多G的话\n说实话我们就算部署起来的话\n速度应该也不会很快\n还有一个原因是\n它官网的版本应该是加了Dspark\n然后进行加速\n但是它开源的版本里面\n我刚才翻了一下它的文件里面好像没有相关的一些描述\n我们再等一等看看三方的会不会把Despark给它接入进去\n它官网的现在这个速度\n我相信它应该是把Despark已经接入进去了\n好的\n如果你对大模型感兴趣的话\n可以关注一下\n然后来在你的API里面切换一下\n其实我上一个视频里面也已经讲到了DeepSick最新的模型\n然后我们来也看一下同步的\n看一下他们最新的官网的一个描述\n我们可以看到它的官网又更新了\n原来右侧的是一个招聘的\n现在不招了\n也不是不招了\n它应该是更新了一个\n然后我们来到它的开放平台上面来看一下\n这个是它的接口文档\n我们重点看它模型的一个更新\n目前在它的接口文档里面还是两个模型\n所以说我们可以看一下模型和价格\n冰干样我记得还有一个\nDipsick Flash, Deep Seek\n他把他的模型,这个4的iflash已经直接切换成这个了\n我们来看一下他的\n模型名称使用Deepseek v4 flash就可以接入到新的模型里面来了\n然后旧的v4 flash和Flash Vision EXP还可以调用\n它应该是已经把它路由到Deepseek v4 e flash这个模型上面来\n然后它的统一名称叫Deepseek v4 flash\n没有V4就是Deepseek flash模型\n新的模型的话你可以直接使用Deepseek v4 flash\n也可以使用Deepseek flash都可以来做接入\n我再给大家带大家来看一下\n我们来到我们的7616里面\n来带大家来做一次配置一下\n其他的所有的三方的工具都是一样的\nOK\n我们在这里面添加一下模型\nOK添加模型\n来做一下检测\n好\n连接失败\n我这个账户应该是欠费的\n正好我今天也开软了一个我自己的一个本地的大模型的一个推比的框架\n然后同时也支持云端\n先简单给大家介绍一下\n我们这里面有对话\n有agent还有视频的通话的\n包括这种可以本地和云端的模式可以跟他先打电话\n但是我这个电脑没在我旁边\n所以说没办法进行沟通\n我们先把它挂了\n回头给单独给大家测试\n然后这一块是生成本地区生成TTS\n然后也可以使用Edge TTS\n也可以使用所有OpenAI兼容的协议的\n来进行本地去生成音频\n这块是生涂的\n然后也可以用本地的MRX生涂\n也可以使用云端的一个生涂\n然后我们可以看一下之前\n历史生涂的一些效果\n这是生涂的\n还有一个OCR\n然后上传文档\n然后进行OCR提取\n这是这个\n除了这个之外\n我们还有翻译\n这是翻译的内容\n中英文翻译\n还有一个提示词\n我们有一个提示词仓库\n然后把它集成进来了\n而这个仓库也已经在\nGitHub已经开完了\n我们在这个里面就是我们的一个模型\n我们可以看一下\nDeepseek v4 flash 还有Deepseek flash\n这个模型都已经可以了\n在我们这个工具里面也可以调用了\n我们来给大家测一下\n来切到Deepseek v4 flash\nDeepseek flash 这个模型\nOK 我们可以看一下\n好\n嗯,抱歉哈\n是这个吧\n这个好像有点小问题\n我再试一下我们的agent\n应该是可以调通的\n是没有问题的\n然后对话应该也是可以的\n我再新建一个对话试一下\n稍等一下\n我们来切换一下\nok 是没有问题的\n那好的 今天就介绍在这儿\n谢谢大家\n如果大家感兴趣的话\n我们这里面的支持的\n我们自己的自带的\n还Deepseek 千问 质朴\nKimi 豆包 文心一言 腾讯\nMinimax 科大讯飞\n和借阅的\n所以这些厂商的大模型在支持\n我们内置了一个agent\n这个agent是py agent\n然后还有一些其他的相关的一些功能\n然后主打的就是本地推理\n那再顺便再多说两句吧\n我们这里面在本地区部署大模型\n支持Nameter CPP VRM和SGLAN\n然后进行部署\n然后在模型库里面\n我们有一些自己挑选的模型\n可以在这里面来进行管理\n然后也可以在这里面进行搜索\n下载GGUF格式的\n然后这个模型也可以在国内进行下载\n我已经做到了加速\n然后我们把本地的这些模型\n启动起来之后我们做了一个统一的网关\n然后接着OpenAPI的一个协议的\nVModel然后Chatter\n然后Response和VMessage和Speech\n然后转翻译的和生图的都有了\n那后面呢我们还会打算把minimax h3接进来\n那好的今天就介绍在这\n谢谢大家", "language": "multilingual", "duration": 647.581313, "segments": [ { "start": 0.0, "end": 1.0, "text": "各位朋友大家好", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 1.0, "end": 3.1, "text": "欢迎来关注昆蓬道今天告诉大家一个好消息", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 3.1, "end": 5.9, "text": "Deepseek v4.1 flash 这个模型正式开源了", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 6.2, "end": 9.3, "text": "那他目前已经把他的模型权重上传到huginface了", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 9.5, "end": 11.4, "text": "然后官方提供的是一个", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 12.7, "end": 15.4, "text": "应该是一个FP8的一个格式的一个模型", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 15.4, "end": 16.8, "text": "模型权重的510个G", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 17.9, "end": 19.6, "text": "我们来看一下他相关的一些参数", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 19.8, "end": 24.1, "text": "那本次他是一个552B的一个大模型", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 24.1, "end": 25.7, "text": "也是一个MOE的架构", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 25.8, "end": 28.8, "text": "然后他本次调整的那个是他的一个CDE的架构", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 28.8, "end": 30.8, "text": "就是由原来的40层的转换器", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 31.0, "end": 34.92, "text": "然后现在优化成了改成20层的英国编码器盒", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 34.92, "end": 36.52, "text": "后面接了20层的解码器", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 36.52, "end": 44.04, "text": "这样的话能够实现前局的KV缓存来进行加速", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 44.36, "end": 49.96, "text": "就是让它能够实现就是预填充的时候", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 49.96, "end": 52.28, "text": "它是一个激活参数是8B的", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 52.28, "end": 56.52, "text": "然后就是解码的时候它的速度是16B的", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 56.52, "end": 60.04, "text": "这样的话又实现了一个全新的一个技术架构", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 60.2, "end": 62.04, "text": "能够实现大模型的一个推理速度加速", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 62.4, "end": 64.8, "text": "然后本次他的这一次的模型", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 64.8, "end": 66.24, "text": "已经全量在他们官网上线了", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 66.24, "end": 68.04, "text": "而且他官网已经把Deepseek v4 Pro", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 68.04, "end": 71.64, "text": "已经路由到他的Deepseek v4 Flash 4.1的模型上面来了", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 71.92, "end": 74.36, "text": "而且他经过他的架构优化", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 74.44, "end": 76.6, "text": "能够让他的kv缓存降低了", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 76.84, "end": 78.64, "text": "就是降低到原来的1%", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 79.8, "end": 85.6, "text": "另外一个就是他用了一个压缩的吸收注意力的算法", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 85.6, "end": 90.96, "text": "然后能够实现也是一个主要是用来加速的", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 90.96, "end": 94.24, "text": "在各个层级用的吸收索引器", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 94.24, "end": 96.48, "text": "然后来进行限制索引", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 96.48, "end": 98.24, "text": "然后实现的一个推理的一个加速", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 98.24, "end": 101.64, "text": "我觉得这块确实在DeepSick在这方面", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 101.96, "end": 103.2, "text": "还是做了很多的一个功夫", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 103.2, "end": 105.28, "text": "另外的一个就是它的一个MHC", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 105.28, "end": 108.32, "text": "这个之前他已经在他的开源的项目里面", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 108.32, "end": 109.0, "text": "已经多次提到过", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 109.0, "end": 112.12, "text": "包括他的Ingram的一个存储器", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 112.12, "end": 113.72, "text": "然后很早的时候应该是在", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 113.72, "end": 115.96, "text": "25年11月份左右的时候", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 115.96, "end": 116.96, "text": "它开源出来的", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 117.72, "end": 120.32, "text": "另外一个就是它本次是一个全新的", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 120.32, "end": 122.72, "text": "完整的多模态的一个大模型", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 122.72, "end": 124.48, "text": "它跟原来Deepseek v4 flash", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 124.48, "end": 126.48, "text": "相当于在前面加了一个补丁包", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 126.48, "end": 128.68, "text": "不一样的是它本次的多模态", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 128.68, "end": 132.04, "text": "是用了Deepseek v8的一种技术架构", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 132.04, "end": 135.36, "text": "然后它是从预训练的时候", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 135.36, "end": 138.28, "text": "就已经开始用到了多模态的一个语料", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 138.28, "end": 139.32, "text": "来进行训练", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 139.32, "end": 142.08, "text": "所以它说它是一个原生的多模态大模型", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 142.08, "end": 143.56, "text": "这个是没有问题的", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 143.56, "end": 147.1, "text": "那我们可以看一下它的一个分数", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 147.1, "end": 149.98, "text": "它本次也很克制", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 149.98, "end": 153.04, "text": "然后它在TerminalBrain Center上面的话", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 153.04, "end": 154.34, "text": "能拿到了一个30分", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 154.34, "end": 157.44, "text": "然后比Kimi5.3都要好一些", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 157.44, "end": 159.94, "text": "然后比Opuls5还是差的很多", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 159.94, "end": 160.54, "text": "我们可以看一下", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 160.54, "end": 162.86, "text": "然后和GPT5.6", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 162.86, "end": 164.1, "text": "那其实GPT6一出来了", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 164.1, "end": 166.54, "text": "所以说还是有很大的一个差距", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 166.54, "end": 168.0, "text": "在国内在TerminalBrain上面", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 168.0, "end": 170.52, "text": "当然这些这叫什么", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 170.52, "end": 174.26, "text": "这些压测的测试的数据集", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 174.26, "end": 176.02, "text": "包括算法都是国外的他们做的", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 176.02, "end": 179.76, "text": "有可能都是由open ai或者说cloud", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 179.76, "end": 180.56, "text": "他们定义的", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 180.56, "end": 181.74, "text": "就谁定义谁说算", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 181.74, "end": 183.02, "text": "他肯定的数据会好一些", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 183.02, "end": 185.18, "text": "当然了这个不能绝对", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 185.18, "end": 187.1, "text": "人家的模型确实我们承认他比较先进", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 187.1, "end": 189.78, "text": "另外一个在这个Deep SWE上面", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 189.78, "end": 191.86, "text": "V1.1的这个版本测试数据以上的话", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 191.86, "end": 193.72, "text": "他应该是拿到了一个最好的一个成绩", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 193.72, "end": 196.82, "text": "然后在安全的这个测评的上面", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 196.82, "end": 197.84, "text": "也拿到一个88分", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 197.84, "end": 201.24, "text": "然后在自动化上面的也能拿到一个54分", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 201.24, "end": 204.24, "text": "目前我看到网上很多的用户评论说", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 204.24, "end": 206.2, "text": "把它接到Depseek里面去目前", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 206.2, "end": 206.48, "text": "不是", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 206.48, "end": 209.52, "text": "接到Codex里面去用Codex Computer Use", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 209.52, "end": 213.14, "text": "是目前性价比和成功率最高的模型之一", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 213.14, "end": 213.88, "text": "除了Depseek", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 213.88, "end": 217.16, "text": "除了GPT官网之外", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 217.16, "end": 221.36, "text": "然后本次算是Depseek真正的机身", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 221.36, "end": 224.0, "text": "然后世界前沿级的大模型", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 224.0, "end": 226.02, "text": "多模态加一照上下文的这个梯堆", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 226.02, "end": 228.66, "text": "目前国内相当于是第三家", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 228.66, "end": 231.16, "text": "第一家相当于是Kimi", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 231.16, "end": 234.02, "text": "第二家相当于是千问", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 234.02, "end": 236.38, "text": "但千问38刚开始推出来的不是多么泰的", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 236.38, "end": 237.0, "text": "我也很奇怪", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 237.0, "end": 237.74, "text": "不知道他现在是不是", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 237.74, "end": 239.88, "text": "我们就把它放在这一列里面", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 239.88, "end": 241.86, "text": "因为我之前说国内只有一家的时候", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 241.86, "end": 243.66, "text": "就有人在评论区说千问也是", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 243.66, "end": 245.7, "text": "我记得千问38刚出来的时候", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 245.7, "end": 246.78, "text": "它是一个文本的多么泰", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 246.78, "end": 248.44, "text": "因为大家都知道", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 248.44, "end": 249.38, "text": "我其实没有做测试", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 249.38, "end": 251.62, "text": "没有深入的测试", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 251.62, "end": 253.72, "text": "然后限制太多了", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 253.72, "end": 254.54, "text": "所以我就放弃了", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 254.54, "end": 258.64, "text": "然后这个是它的整个的一个测评的一个分数", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 258.64, "end": 260.26, "text": "比如说在各个领域", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 260.26, "end": 261.8, "text": "然后还是非常亮眼的", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 261.8, "end": 262.7, "text": "这个黑色的就表示", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 262.7, "end": 265.5, "text": "它是目前的在已知的测评的这些里面", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 265.5, "end": 266.28, "text": "效果是", "language": "zh", "chunk": 0, "local_language_run": 0, "language_run": 0 }, { "start": 268.558, "end": 269.408, "text": "效果是最好的", "language": "zh", "chunk": 0, "local_language_run": 1, "language_run": 1 }, { "start": 270.308, "end": 271.648, "text": "那我们可以看一下", "language": "zh", "chunk": 0, "local_language_run": 1, "language_run": 1 }, { "start": 271.648, "end": 274.108, "text": "这个是它的一个base模型和它的", "language": "zh", "chunk": 0, "local_language_run": 1, "language_run": 1 }, { "start": 274.108, "end": 275.248, "text": "对就是这个模型", "language": "zh", "chunk": 0, "local_language_run": 1, "language_run": 1 }, { "start": 276.348, "end": 280.808, "text": "它在很多的参数已经非常接近v4Pro", "language": "zh", "chunk": 0, "local_language_run": 1, "language_run": 1 }, { "start": 280.808, "end": 282.948, "text": "和已经超过的Pro的模型", "language": "zh", "chunk": 0, "local_language_run": 1, "language_run": 1 }, { "start": 282.948, "end": 284.348, "text": "所以说他们官网已经把这个", "language": "zh", "chunk": 0, "local_language_run": 1, "language_run": 1 }, { "start": 284.348, "end": 287.048, "text": "已经路由到Pro路由到这个模型上面来", "language": "zh", "chunk": 0, "local_language_run": 1, "language_run": 1 }, { "start": 287.048, "end": 288.908, "text": "第一个是为了降低它的一个推理成本", "language": "zh", "chunk": 0, "local_language_run": 1, "language_run": 1 }, { "start": 288.908, "end": 293.008, "text": "第二个也是因为这个模型的能力确实提升非常明显", "language": "zh", "chunk": 0, "local_language_run": 1, "language_run": 1 }, { "start": 293.008, "end": 295.808, "text": "还有一个的最主要原因是网上被很多用户吐槽", "language": "zh", "chunk": 0, "local_language_run": 1, "language_run": 1 }, { "start": 295.808, "end": 297.408, "text": "Deepstick v4Pro的幻觉比较严重", "language": "zh", "chunk": 0, "local_language_run": 1, "language_run": 1 }, { "start": 297.408, "end": 300.588, "text": "实际上我使用下来整体感觉还是非常不错的", "language": "zh", "chunk": 0, "local_language_run": 1, "language_run": 1 }, { "start": 301.168, "end": 303.768, "text": "他这些测评的分数我们就不一一讲了", "language": "zh", "chunk": 0, "local_language_run": 1, "language_run": 1 }, { "start": 303.768, "end": 305.268, "text": "因为刚才我们已经带大家看了一下", "language": "zh", "chunk": 0, "local_language_run": 1, "language_run": 1 }, { "start": 305.268, "end": 307.308, "text": "他跟国外的这些模型的做一些对比", "language": "zh", "chunk": 0, "local_language_run": 1, "language_run": 1 }, { "start": 307.728, "end": 310.048, "text": "其实我们可以看到目前他应该是", "language": "zh", "chunk": 0, "local_language_run": 1, "language_run": 1 }, { "start": 310.048, "end": 313.668, "text": "激升到了国内的大模型的地体对", "language": "zh", "chunk": 0, "local_language_run": 1, "language_run": 1 }, { "start": 313.668, "end": 314.228, "text": "毫无疑问", "language": "zh", "chunk": 0, "local_language_run": 1, "language_run": 1 }, { "start": 314.228, "end": 317.148, "text": "所以说DeepSick还是非常强的", "language": "zh", "chunk": 0, "local_language_run": 1, "language_run": 1 }, { "start": 317.148, "end": 322.668, "text": "目前他官方还没有去推出", "language": "zh", "chunk": 0, "local_language_run": 1, "language_run": 1 }, { "start": 322.668, "end": 324.408, "text": "然后VM和SGLAN部署的", "language": "zh", "chunk": 0, "local_language_run": 1, "language_run": 1 }, { "start": 324.408, "end": 325.668, "text": "然后我相信应该很快", "language": "zh", "chunk": 0, "local_language_run": 1, "language_run": 1 }, { "start": 325.668, "end": 327.628, "text": "VMHSG浪就有对您的支持", "language": "zh", "chunk": 0, "local_language_run": 1, "language_run": 1 }, { "start": 327.628, "end": 329.308, "text": "就算没有对您也有对E的支持", "language": "zh", "chunk": 0, "local_language_run": 1, "language_run": 1 }, { "start": 329.308, "end": 332.388, "text": "刚才我们在一个我们私有化部署大木型的群里", "language": "zh", "chunk": 0, "local_language_run": 1, "language_run": 1 }, { "start": 332.388, "end": 335.348, "text": "就有人说今天晚上想蹲一个40E的本地部署", "language": "zh", "chunk": 0, "local_language_run": 1, "language_run": 1 }, { "start": 335.348, "end": 336.048, "text": "我感觉够呛", "language": "zh", "chunk": 0, "local_language_run": 1, "language_run": 1 }, { "start": 336.048, "end": 337.728, "text": "其实我们手上也有机器", "language": "zh", "chunk": 0, "local_language_run": 1, "language_run": 1 }, { "start": 337.728, "end": 339.888, "text": "但是不打算部署了", "language": "zh", "chunk": 0, "local_language_run": 1, "language_run": 1 }, { "start": 339.888, "end": 342.688, "text": "因为它500多G的话", "language": "zh", "chunk": 0, "local_language_run": 1, "language_run": 1 }, { "start": 342.688, "end": 344.348, "text": "说实话我们就算部署起来的话", "language": "zh", "chunk": 0, "local_language_run": 1, "language_run": 1 }, { "start": 344.348, "end": 346.648, "text": "速度应该也不会很快", "language": "zh", "chunk": 0, "local_language_run": 1, "language_run": 1 }, { "start": 346.648, "end": 348.108, "text": "还有一个原因是", "language": "zh", "chunk": 0, "local_language_run": 1, "language_run": 1 }, { "start": 348.108, "end": 351.548, "text": "它官网的版本应该是加了Dspark", "language": "zh", "chunk": 0, "local_language_run": 1, "language_run": 1 }, { "start": 351.548, "end": 352.508, "text": "然后进行加速", "language": "zh", "chunk": 0, "local_language_run": 1, "language_run": 1 }, { "start": 352.508, "end": 353.788, "text": "但是它开源的版本里面", "language": "zh", "chunk": 0, "local_language_run": 1, "language_run": 1 }, { "start": 353.788, "end": 357.188, "text": "我刚才翻了一下它的文件里面好像没有相关的一些描述", "language": "zh", "chunk": 0, "local_language_run": 1, "language_run": 1 }, { "start": 357.188, "end": 361.708, "text": "我们再等一等看看三方的会不会把Despark给它接入进去", "language": "zh", "chunk": 0, "local_language_run": 1, "language_run": 1 }, { "start": 361.708, "end": 362.908, "text": "它官网的现在这个速度", "language": "zh", "chunk": 0, "local_language_run": 1, "language_run": 1 }, { "start": 362.908, "end": 365.788, "text": "我相信它应该是把Despark已经接入进去了", "language": "zh", "chunk": 0, "local_language_run": 1, "language_run": 1 }, { "start": 365.788, "end": 367.748, "text": "好的", "language": "zh", "chunk": 0, "local_language_run": 1, "language_run": 1 }, { "start": 367.748, "end": 369.868, "text": "如果你对大模型感兴趣的话", "language": "zh", "chunk": 0, "local_language_run": 1, "language_run": 1 }, { "start": 369.868, "end": 370.628, "text": "可以关注一下", "language": "zh", "chunk": 0, "local_language_run": 1, "language_run": 1 }, { "start": 370.628, "end": 375.808, "text": "然后来在你的API里面切换一下", "language": "zh", "chunk": 0, "local_language_run": 1, "language_run": 1 }, { "start": 375.808, "end": 379.108, "text": "其实我上一个视频里面也已经讲到了DeepSick最新的模型", "language": "zh", "chunk": 0, "local_language_run": 1, "language_run": 1 }, { "start": 379.108, "end": 380.708, "text": "然后我们来也看一下同步的", "language": "zh", "chunk": 0, "local_language_run": 1, "language_run": 1 }, { "start": 380.708, "end": 383.028, "text": "看一下他们最新的官网的一个描述", "language": "zh", "chunk": 0, "local_language_run": 1, "language_run": 1 }, { "start": 385.292, "end": 386.882, "text": "我们可以看到它的官网又更新了", "language": "zh", "chunk": 0, "local_language_run": 2, "language_run": 2 }, { "start": 386.882, "end": 388.162, "text": "原来右侧的是一个招聘的", "language": "zh", "chunk": 0, "local_language_run": 2, "language_run": 2 }, { "start": 388.162, "end": 388.902, "text": "现在不招了", "language": "zh", "chunk": 0, "local_language_run": 2, "language_run": 2 }, { "start": 388.902, "end": 389.582, "text": "也不是不招了", "language": "zh", "chunk": 0, "local_language_run": 2, "language_run": 2 }, { "start": 389.582, "end": 391.062, "text": "它应该是更新了一个", "language": "zh", "chunk": 0, "local_language_run": 2, "language_run": 2 }, { "start": 391.062, "end": 394.182, "text": "然后我们来到它的开放平台上面来看一下", "language": "zh", "chunk": 0, "local_language_run": 2, "language_run": 2 }, { "start": 394.182, "end": 396.762, "text": "这个是它的接口文档", "language": "zh", "chunk": 0, "local_language_run": 2, "language_run": 2 }, { "start": 396.762, "end": 398.942, "text": "我们重点看它模型的一个更新", "language": "zh", "chunk": 0, "local_language_run": 2, "language_run": 2 }, { "start": 398.942, "end": 402.442, "text": "目前在它的接口文档里面还是两个模型", "language": "zh", "chunk": 0, "local_language_run": 2, "language_run": 2 }, { "start": 402.442, "end": 405.942, "text": "所以说我们可以看一下模型和价格", "language": "zh", "chunk": 0, "local_language_run": 2, "language_run": 2 }, { "start": 407.98, "end": 409.145, "text": "冰干样我记得还有一个", "language": "zh", "chunk": 0, "local_language_run": 3, "language_run": 3 }, { "start": 412.295, "end": 413.796, "text": "Dipsick Flash, Deep Seek", "language": "zh", "chunk": 0, "local_language_run": 4, "language_run": 4 }, { "start": 417.905, "end": 421.955, "text": "他把他的模型,这个4的iflash已经直接切换成这个了", "language": "zh", "chunk": 0, "local_language_run": 5, "language_run": 5 }, { "start": 421.955, "end": 423.695, "text": "我们来看一下他的", "language": "zh", "chunk": 0, "local_language_run": 5, "language_run": 5 }, { "start": 426.656, "end": 429.806, "text": "模型名称使用Deepseek v4 flash就可以接入到新的模型里面来了", "language": "zh", "chunk": 0, "local_language_run": 6, "language_run": 6 }, { "start": 429.806, "end": 434.186, "text": "然后旧的v4 flash和Flash Vision EXP还可以调用", "language": "zh", "chunk": 0, "local_language_run": 6, "language_run": 6 }, { "start": 434.186, "end": 437.806, "text": "它应该是已经把它路由到Deepseek v4 e flash这个模型上面来", "language": "zh", "chunk": 0, "local_language_run": 6, "language_run": 6 }, { "start": 437.806, "end": 439.766, "text": "然后它的统一名称叫Deepseek v4 flash", "language": "zh", "chunk": 0, "local_language_run": 6, "language_run": 6 }, { "start": 439.766, "end": 442.186, "text": "没有V4就是Deepseek flash模型", "language": "zh", "chunk": 0, "local_language_run": 6, "language_run": 6 }, { "start": 442.186, "end": 445.246, "text": "新的模型的话你可以直接使用Deepseek v4 flash", "language": "zh", "chunk": 0, "local_language_run": 6, "language_run": 6 }, { "start": 445.246, "end": 447.966, "text": "也可以使用Deepseek flash都可以来做接入", "language": "zh", "chunk": 0, "local_language_run": 6, "language_run": 6 }, { "start": 447.966, "end": 450.466, "text": "我再给大家带大家来看一下", "language": "zh", "chunk": 0, "local_language_run": 6, "language_run": 6 }, { "start": 450.466, "end": 453.066, "text": "我们来到我们的7616里面", "language": "zh", "chunk": 0, "local_language_run": 6, "language_run": 6 }, { "start": 453.066, "end": 455.766, "text": "来带大家来做一次配置一下", "language": "zh", "chunk": 0, "local_language_run": 6, "language_run": 6 }, { "start": 455.766, "end": 458.366, "text": "其他的所有的三方的工具都是一样的", "language": "zh", "chunk": 0, "local_language_run": 6, "language_run": 6 }, { "start": 460.766, "end": 461.566, "text": "OK", "language": "zh", "chunk": 0, "local_language_run": 6, "language_run": 7 }, { "start": 463.722, "end": 465.572, "text": "我们在这里面添加一下模型", "language": "zh", "chunk": 1, "local_language_run": 1, "language_run": 8 }, { "start": 465.572, "end": 466.832, "text": "OK添加模型", "language": "zh", "chunk": 1, "local_language_run": 1, "language_run": 8 }, { "start": 466.832, "end": 468.032, "text": "来做一下检测", "language": "zh", "chunk": 1, "local_language_run": 1, "language_run": 8 }, { "start": 469.332, "end": 470.982, "text": "好", "language": "zh", "chunk": 1, "local_language_run": 2, "language_run": 8 }, { "start": 472.856, "end": 473.506, "text": "连接失败", "language": "zh", "chunk": 1, "local_language_run": 3, "language_run": 8 }, { "start": 473.506, "end": 476.506, "text": "我这个账户应该是欠费的", "language": "zh", "chunk": 1, "local_language_run": 3, "language_run": 8 }, { "start": 476.506, "end": 484.906, "text": "正好我今天也开软了一个我自己的一个本地的大模型的一个推比的框架", "language": "zh", "chunk": 1, "local_language_run": 3, "language_run": 8 }, { "start": 484.906, "end": 486.146, "text": "然后同时也支持云端", "language": "zh", "chunk": 1, "local_language_run": 3, "language_run": 8 }, { "start": 486.146, "end": 487.226, "text": "先简单给大家介绍一下", "language": "zh", "chunk": 1, "local_language_run": 3, "language_run": 8 }, { "start": 487.226, "end": 488.626, "text": "我们这里面有对话", "language": "zh", "chunk": 1, "local_language_run": 3, "language_run": 8 }, { "start": 488.626, "end": 490.386, "text": "有agent还有视频的通话的", "language": "zh", "chunk": 1, "local_language_run": 3, "language_run": 8 }, { "start": 490.386, "end": 493.586, "text": "包括这种可以本地和云端的模式可以跟他先打电话", "language": "zh", "chunk": 1, "local_language_run": 3, "language_run": 8 }, { "start": 493.586, "end": 494.786, "text": "但是我这个电脑没在我旁边", "language": "zh", "chunk": 1, "local_language_run": 3, "language_run": 8 }, { "start": 494.786, "end": 495.906, "text": "所以说没办法进行沟通", "language": "zh", "chunk": 1, "local_language_run": 3, "language_run": 8 }, { "start": 495.906, "end": 496.626, "text": "我们先把它挂了", "language": "zh", "chunk": 1, "local_language_run": 3, "language_run": 8 }, { "start": 496.626, "end": 497.906, "text": "回头给单独给大家测试", "language": "zh", "chunk": 1, "local_language_run": 3, "language_run": 8 }, { "start": 497.906, "end": 501.426, "text": "然后这一块是生成本地区生成TTS", "language": "zh", "chunk": 1, "local_language_run": 3, "language_run": 8 }, { "start": 501.426, "end": 502.866, "text": "然后也可以使用Edge TTS", "language": "zh", "chunk": 1, "local_language_run": 3, "language_run": 8 }, { "start": 502.866, "end": 504.866, "text": "也可以使用所有OpenAI兼容的协议的", "language": "zh", "chunk": 1, "local_language_run": 3, "language_run": 8 }, { "start": 504.866, "end": 506.386, "text": "来进行本地去生成音频", "language": "zh", "chunk": 1, "local_language_run": 3, "language_run": 8 }, { "start": 506.866, "end": 507.986, "text": "这块是生涂的", "language": "zh", "chunk": 1, "local_language_run": 3, "language_run": 8 }, { "start": 507.986, "end": 510.466, "text": "然后也可以用本地的MRX生涂", "language": "zh", "chunk": 1, "local_language_run": 3, "language_run": 8 }, { "start": 510.466, "end": 511.826, "text": "也可以使用云端的一个生涂", "language": "zh", "chunk": 1, "local_language_run": 3, "language_run": 8 }, { "start": 511.826, "end": 512.866, "text": "然后我们可以看一下之前", "language": "zh", "chunk": 1, "local_language_run": 3, "language_run": 8 }, { "start": 513.226, "end": 514.466, "text": "历史生涂的一些效果", "language": "zh", "chunk": 1, "local_language_run": 3, "language_run": 8 }, { "start": 514.866, "end": 515.506, "text": "这是生涂的", "language": "zh", "chunk": 1, "local_language_run": 3, "language_run": 8 }, { "start": 515.826, "end": 516.546, "text": "还有一个OCR", "language": "zh", "chunk": 1, "local_language_run": 3, "language_run": 8 }, { "start": 516.546, "end": 517.666, "text": "然后上传文档", "language": "zh", "chunk": 1, "local_language_run": 3, "language_run": 8 }, { "start": 517.666, "end": 518.866, "text": "然后进行OCR提取", "language": "zh", "chunk": 1, "local_language_run": 3, "language_run": 8 }, { "start": 519.026, "end": 519.506, "text": "这是这个", "language": "zh", "chunk": 1, "local_language_run": 3, "language_run": 8 }, { "start": 519.866, "end": 520.426, "text": "除了这个之外", "language": "zh", "chunk": 1, "local_language_run": 3, "language_run": 8 }, { "start": 520.426, "end": 521.506, "text": "我们还有翻译", "language": "zh", "chunk": 1, "local_language_run": 3, "language_run": 8 }, { "start": 521.746, "end": 522.706, "text": "这是翻译的内容", "language": "zh", "chunk": 1, "local_language_run": 3, "language_run": 8 }, { "start": 522.706, "end": 523.506, "text": "中英文翻译", "language": "zh", "chunk": 1, "local_language_run": 3, "language_run": 8 }, { "start": 523.826, "end": 525.106, "text": "还有一个提示词", "language": "zh", "chunk": 1, "local_language_run": 3, "language_run": 8 }, { "start": 525.346, "end": 526.386, "text": "我们有一个提示词仓库", "language": "zh", "chunk": 1, "local_language_run": 3, "language_run": 8 }, { "start": 526.386, "end": 527.266, "text": "然后把它集成进来了", "language": "zh", "chunk": 1, "local_language_run": 3, "language_run": 8 }, { "start": 527.586, "end": 528.786, "text": "而这个仓库也已经在", "language": "zh", "chunk": 1, "local_language_run": 3, "language_run": 8 }, { "start": 531.426, "end": 532.426, "text": "GitHub已经开完了", "language": "zh", "chunk": 1, "local_language_run": 3, "language_run": 9 }, { "start": 532.426, "end": 535.426, "text": "我们在这个里面就是我们的一个模型", "language": "zh", "chunk": 1, "local_language_run": 3, "language_run": 9 }, { "start": 535.426, "end": 536.186, "text": "我们可以看一下", "language": "zh", "chunk": 1, "local_language_run": 3, "language_run": 9 }, { "start": 536.186, "end": 538.386, "text": "Deepseek v4 flash 还有Deepseek flash", "language": "zh", "chunk": 1, "local_language_run": 3, "language_run": 9 }, { "start": 538.386, "end": 539.986, "text": "这个模型都已经可以了", "language": "zh", "chunk": 1, "local_language_run": 3, "language_run": 9 }, { "start": 539.986, "end": 541.866, "text": "在我们这个工具里面也可以调用了", "language": "zh", "chunk": 1, "local_language_run": 3, "language_run": 9 }, { "start": 541.866, "end": 542.826, "text": "我们来给大家测一下", "language": "zh", "chunk": 1, "local_language_run": 3, "language_run": 9 }, { "start": 543.066, "end": 545.466, "text": "来切到Deepseek v4 flash", "language": "zh", "chunk": 1, "local_language_run": 3, "language_run": 9 }, { "start": 545.826, "end": 547.226, "text": "Deepseek flash 这个模型", "language": "zh", "chunk": 1, "local_language_run": 3, "language_run": 9 }, { "start": 547.226, "end": 548.497, "text": "OK 我们可以看一下", "language": "zh", "chunk": 1, "local_language_run": 3, "language_run": 9 }, { "start": 549.769, "end": 551.061, "text": "好", "language": "zh", "chunk": 1, "local_language_run": 4, "language_run": 9 }, { "start": 552.75, "end": 554.4, "text": "嗯,抱歉哈", "language": "zh", "chunk": 1, "local_language_run": 5, "language_run": 9 }, { "start": 558.031, "end": 558.681, "text": "是这个吧", "language": "zh", "chunk": 1, "local_language_run": 6, "language_run": 10 }, { "start": 564.952, "end": 568.382, "text": "这个好像有点小问题", "language": "zh", "chunk": 1, "local_language_run": 7, "language_run": 11 }, { "start": 568.382, "end": 569.702, "text": "我再试一下我们的agent", "language": "zh", "chunk": 1, "local_language_run": 7, "language_run": 11 }, { "start": 569.702, "end": 573.802, "text": "应该是可以调通的", "language": "zh", "chunk": 1, "local_language_run": 7, "language_run": 11 }, { "start": 573.802, "end": 575.362, "text": "是没有问题的", "language": "zh", "chunk": 1, "local_language_run": 7, "language_run": 11 }, { "start": 575.362, "end": 576.902, "text": "然后对话应该也是可以的", "language": "zh", "chunk": 1, "local_language_run": 7, "language_run": 11 }, { "start": 576.902, "end": 578.062, "text": "我再新建一个对话试一下", "language": "zh", "chunk": 1, "local_language_run": 7, "language_run": 11 }, { "start": 578.062, "end": 578.542, "text": "稍等一下", "language": "zh", "chunk": 1, "local_language_run": 7, "language_run": 11 }, { "start": 578.542, "end": 580.822, "text": "我们来切换一下", "language": "zh", "chunk": 1, "local_language_run": 7, "language_run": 11 }, { "start": 583.992, "end": 585.402, "text": "ok 是没有问题的", "language": "zh", "chunk": 1, "local_language_run": 8, "language_run": 12 }, { "start": 585.402, "end": 588.762, "text": "那好的 今天就介绍在这儿", "language": "zh", "chunk": 1, "local_language_run": 8, "language_run": 12 }, { "start": 588.762, "end": 589.242, "text": "谢谢大家", "language": "zh", "chunk": 1, "local_language_run": 8, "language_run": 12 }, { "start": 589.242, "end": 590.202, "text": "如果大家感兴趣的话", "language": "zh", "chunk": 1, "local_language_run": 8, "language_run": 12 }, { "start": 590.202, "end": 591.242, "text": "我们这里面的支持的", "language": "zh", "chunk": 1, "local_language_run": 8, "language_run": 12 }, { "start": 591.242, "end": 592.322, "text": "我们自己的自带的", "language": "zh", "chunk": 1, "local_language_run": 8, "language_run": 12 }, { "start": 592.322, "end": 593.882, "text": "还Deepseek 千问 质朴", "language": "zh", "chunk": 1, "local_language_run": 8, "language_run": 12 }, { "start": 593.882, "end": 596.042, "text": "Kimi 豆包 文心一言 腾讯", "language": "zh", "chunk": 1, "local_language_run": 8, "language_run": 12 }, { "start": 596.042, "end": 598.002, "text": "Minimax 科大讯飞", "language": "zh", "chunk": 1, "local_language_run": 8, "language_run": 12 }, { "start": 598.002, "end": 598.722, "text": "和借阅的", "language": "zh", "chunk": 1, "local_language_run": 8, "language_run": 12 }, { "start": 598.722, "end": 601.122, "text": "所以这些厂商的大模型在支持", "language": "zh", "chunk": 1, "local_language_run": 8, "language_run": 12 }, { "start": 601.122, "end": 602.442, "text": "我们内置了一个agent", "language": "zh", "chunk": 1, "local_language_run": 8, "language_run": 12 }, { "start": 602.442, "end": 603.642, "text": "这个agent是py agent", "language": "zh", "chunk": 1, "local_language_run": 8, "language_run": 12 }, { "start": 603.642, "end": 606.362, "text": "然后还有一些其他的相关的一些功能", "language": "zh", "chunk": 1, "local_language_run": 8, "language_run": 12 }, { "start": 606.362, "end": 608.002, "text": "然后主打的就是本地推理", "language": "zh", "chunk": 1, "local_language_run": 8, "language_run": 12 }, { "start": 608.002, "end": 610.402, "text": "那再顺便再多说两句吧", "language": "zh", "chunk": 1, "local_language_run": 8, "language_run": 12 }, { "start": 610.402, "end": 612.282, "text": "我们这里面在本地区部署大模型", "language": "zh", "chunk": 1, "local_language_run": 8, "language_run": 12 }, { "start": 612.282, "end": 614.142, "text": "支持Nameter CPP VRM和SGLAN", "language": "zh", "chunk": 1, "local_language_run": 8, "language_run": 12 }, { "start": 614.142, "end": 615.042, "text": "然后进行部署", "language": "zh", "chunk": 1, "local_language_run": 8, "language_run": 12 }, { "start": 615.042, "end": 616.282, "text": "然后在模型库里面", "language": "zh", "chunk": 1, "local_language_run": 8, "language_run": 12 }, { "start": 616.282, "end": 618.682, "text": "我们有一些自己挑选的模型", "language": "zh", "chunk": 1, "local_language_run": 8, "language_run": 12 }, { "start": 618.682, "end": 619.962, "text": "可以在这里面来进行管理", "language": "zh", "chunk": 1, "local_language_run": 8, "language_run": 12 }, { "start": 619.962, "end": 621.322, "text": "然后也可以在这里面进行搜索", "language": "zh", "chunk": 1, "local_language_run": 8, "language_run": 12 }, { "start": 621.322, "end": 622.522, "text": "下载GGUF格式的", "language": "zh", "chunk": 1, "local_language_run": 8, "language_run": 12 }, { "start": 623.162, "end": 626.602, "text": "然后这个模型也可以在国内进行下载", "language": "zh", "chunk": 1, "local_language_run": 8, "language_run": 12 }, { "start": 626.602, "end": 627.602, "text": "我已经做到了加速", "language": "zh", "chunk": 1, "local_language_run": 8, "language_run": 12 }, { "start": 627.602, "end": 629.402, "text": "然后我们把本地的这些模型", "language": "zh", "chunk": 1, "local_language_run": 8, "language_run": 12 }, { "start": 629.402, "end": 631.322, "text": "启动起来之后我们做了一个统一的网关", "language": "zh", "chunk": 1, "local_language_run": 8, "language_run": 12 }, { "start": 631.322, "end": 633.522, "text": "然后接着OpenAPI的一个协议的", "language": "zh", "chunk": 1, "local_language_run": 8, "language_run": 12 }, { "start": 633.522, "end": 634.802, "text": "VModel然后Chatter", "language": "zh", "chunk": 1, "local_language_run": 8, "language_run": 12 }, { "start": 634.802, "end": 637.482, "text": "然后Response和VMessage和Speech", "language": "zh", "chunk": 1, "local_language_run": 8, "language_run": 12 }, { "start": 637.482, "end": 641.042, "text": "然后转翻译的和生图的都有了", "language": "zh", "chunk": 1, "local_language_run": 8, "language_run": 12 }, { "start": 641.042, "end": 644.002, "text": "那后面呢我们还会打算把minimax h3接进来", "language": "zh", "chunk": 1, "local_language_run": 8, "language_run": 12 }, { "start": 644.002, "end": 645.602, "text": "那好的今天就介绍在这", "language": "zh", "chunk": 1, "local_language_run": 8, "language_run": 12 }, { "start": 645.602, "end": 645.994, "text": "谢谢大家", "language": "zh", "chunk": 1, "local_language_run": 8, "language_run": 12 } ], "segmented_multilingual_transcription": { "status": "STABLE_SEGMENTED_MULTILINGUAL_REVIEW", "created_at": "2026-09-11T08:50:37", "manifest": "/Users/sagawa/AI_WORK/video_notes/url/20260911_084209__DeepSeek_V4.1_Flash开源了_百万上下文_视觉理解_超长输出_新模型亮点与API价格一次看懂/segmented_work/chunk_manifest.json", "audio": "/Users/sagawa/AI_WORK/video_notes/url/20260911_084209__DeepSeek_V4.1_Flash开源了_百万上下文_视觉理解_超长输出_新模型亮点与API价格一次看懂/audio_16k_mono.wav", "audio_sha256": "a4e23a53de8925f3599b3bb156c8a3addcbcee5872a6d0fafe8e667584a5adfa", "audio_duration": 647.581313, "model": "/Users/sagawa/AI_WORK/cache/huggingface/hub/models--mlx-community--whisper-large-v3-turbo/snapshots/a4aaeec0636e6fef84abdcbe3544cb2bf7e9f6fb", "language_profile": "/Users/sagawa/AI_WORK/video_notes/url/20260911_084209__DeepSeek_V4.1_Flash开源了_百万上下文_视觉理解_超长输出_新模型亮点与API价格一次看懂/multilingual_work/multilingual_language_profile.json", "dominant_languages": [ "zh" ], "suppressed_languages": [ "en", "es", "de", "ko" ], "chunk_count": 2, "completed_chunks": 2, "resumed_chunks": 0, "merged_segment_count": 278, "detected_languages": [ "zh" ], "transcript_languages": [ "zh" ], "missing_languages": [], "unexpected_languages": [], "language_sequence": [ "zh" ], "speech_region_seconds": 617.93, "speech_covered_seconds": 588.271, "speech_coverage_ratio": 0.9520026540223002, "tail_repeat_run": 1, "tail_repeat_text": "各位朋友大家好", "chunk_needs_review": [ { "chunk": 0, "runs": [ 4 ], "exit_code": 43 } ], "qc_ok": false, "qc_errors": [ "one or more chunks require review" ], "complete_review_artifact": true } }