{ "text": "各位朋友大家好。歡迎大家關注昆蓬道。今天告訴大家一個好消息。Deepseek V4.1 Flash這個模型正式開源了。\n那它目前已經把它的模型全中上傳到Hugging Face了。然後官方提供的是一個應該是一個FP8的格式的一個模型。\n模型全中的510個G。我們來看一下它相關的一些參數。那本次它是一個552B的一個大模型。也是一個MOE的架構。\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层的这个转换器\n然后现在优化成了改成20层的英国编码器和后面接了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就是让他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然后本次它的这一次的模型\n已经全量在他们官网上线了\n而且它官网已经把Deepseek v4 Pro\n已经路由到它的Deepseek v4 Flash 4.1的模型上面来了\n而且它经过它的架构优化\n能够让它的kb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另外的一个就是他的一个MHC\n这个之前他已经在他的开源的项目里面\n也多次提到过\n包括他的enggram的一个存储器\n然后很早的时候\n应该是在25年11月份左右的时候\n他开源出来的\n另外一个就是他本次是一个全新的\n完整的多模态的一个大模型\n他跟原来Deepseek v4 flash\n相当于在前面加了一个补丁包\n不一样的是他本次的多模态\n是用了Deepseek v4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谋态的一个预料\n来进行训练\n所以他说它是一个原生的多谋态大模型\n这个是没有问题的\n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Terminal Branch 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Opus5还是差的很多\n我们可以看一下\n然后和GPT5.6\n其实GPT6一出来了\n所以说还是有很大的一个差距的\n在国内在Terminibank上面\n当然这些\n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OpenAI或者说Cloud他们定义的\n就谁定义谁说算嘛\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Deep SWE上面\nv1.1的版本测试数据以上的话\n他应该是拿到了一个最好的成绩\n然后在安全的测评的上面也拿到了一个88分\n然后在自动化上面也拿到了一个54分\n目前我看到网上很多的用户评论说\n把它接到Deepseek里面去\n目前接到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Deepseek\n除了GPT官网之外\n然后本次算是Deep Seek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态这个梯堆\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模态的\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它是一个文本的多模态\n因为大家都知道我其实没有做测试\n就是没有深入的测试\n卖Token Plan\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效果是最好的\n我们可以看一下\n这个是它的一个base模型\n和它的对就是这个模型\n它在很多的参数已经非常接近V4 Pro和已经超过的Pro的模型\n所以说他们官网已经把这个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 V4 Pro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这些测评的分数我们就不一一讲\n因为刚才我们已经带大家看了一下\n它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它应该是\n牺牲到了国内的大模型的第一体队\n毫无疑问\n所以说DeepSick还是非常强的\n目前它官方还没有去推出\n然后VM和SGLAN部署的\n然后我相信应该很快\nVM和SGLAN就有对您的支持\n就算没有对您也有对一的支持\n刚才我们在一个我们私有化\n部署大模型的群里面就有人说\n今天晚上想蹲一个4.1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\n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\n三方会不会把Dspark给它接入进去\n他官网的现在这个速度\n我相信他应该是把despark已经接入进去了\n那好的\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\n就是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我们可以看到他的官网又更新了\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他应该是更新了一个\n然后我们来到它的开放平台上\n我们来看一下\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\n还是两个模型\n所以说我们可以看一下模型和价格\n冰干压我记得还有一个\nDipsick Flash\nDipsick V4 Pro\n它把它的模型\n这个4.1 Flash\n已经直接切换成这个了\n我们来看一下它的\n模型名称使用Deepseek V4 Flash\n就可以接入到新的模型里面来了\n然后旧的V4 Flash和Flash Vision XP\n还可以调用\n它应该是已经把它路由到\nDeepseek 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\n你可以直接使用Deepseek V4 Flash\n也可以使用Deepseek Flash\n都可以来做我介入\n我再给大家带大家来看一下\n我们来到我们的7166里面\n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\nOK。我們在這裡面添加一下模型。OK。添加模型。來做一下檢測。\n连接失败\n我这个账户应该是欠费的\n正好我今天也开玩了一个\n我自己的一个本地的大模型的一个推理的框架\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\n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去生成TTS\n然后可以使用Edge TTS\n也可以使用所谓OpenAI兼容的协议的\n来进行本地去生成音频\n这块是生图的\n然后也可以用本地的MRX生图\n也可以使用云端的一个生图\n然后我们可以看一下之前历史生图的一些效果\n这是生图的\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GitHub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不欠\n是这个吧\n我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然后这个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\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\nBinimax 科大讯飞和借阅的\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GGOF格式的\n然后这个模型也可以在国内进行下载\n我已经做到加速\n然后我们把本地的这些模型\n启动起来之后\n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后面我们还会打算把minimax H3接进来\n好的\n今天就介绍在这\n谢谢大家", "language": "zh", "duration": 647.581313, "segments": [ { "start": 0.0, "end": 6.0, "text": "各位朋友大家好。歡迎大家關注昆蓬道。今天告訴大家一個好消息。Deepseek V4.1 Flash這個模型正式開源了。", "chunk": 0, "language": "zh", "avg_logprob": -0.273947292700746, "no_speech_prob": 9.472454938486141e-12, "compression_ratio": 1.334384858044164 }, { "start": 6.0, "end": 15.0, "text": "那它目前已經把它的模型全中上傳到Hugging Face了。然後官方提供的是一個應該是一個FP8的格式的一個模型。", "chunk": 0, "language": "zh", "avg_logprob": -0.273947292700746, "no_speech_prob": 9.472454938486141e-12, "compression_ratio": 1.334384858044164 }, { "start": 15.0, "end": 26.0, "text": "模型全中的510個G。我們來看一下它相關的一些參數。那本次它是一個552B的一個大模型。也是一個MOE的架構。", "chunk": 0, "language": "zh", "avg_logprob": -0.273947292700746, "no_speech_prob": 9.472454938486141e-12, "compression_ratio": 1.334384858044164 }, { "start": 26.0, "end": 28.76, "text": "然后他本次调整的那个是他的一个CDE的架构", "chunk": 0, "language": "zh", "avg_logprob": -0.15405842707707332, "no_speech_prob": 2.06851133444097e-11, "compression_ratio": 1.291497975708502 }, { "start": 28.76, "end": 30.68, "text": "就是由原来的40层的这个转换器", "chunk": 0, "language": "zh", "avg_logprob": -0.15405842707707332, "no_speech_prob": 2.06851133444097e-11, "compression_ratio": 1.291497975708502 }, { "start": 30.68, "end": 36.5, "text": "然后现在优化成了改成20层的英国编码器和后面接了20层的解码器", "chunk": 0, "language": "zh", "avg_logprob": -0.15405842707707332, "no_speech_prob": 2.06851133444097e-11, "compression_ratio": 1.291497975708502 }, { "start": 36.5, "end": 43.9, "text": "这样的话能够实现前局的KV缓存来进行加速", "chunk": 0, "language": "zh", "avg_logprob": -0.15405842707707332, "no_speech_prob": 2.06851133444097e-11, "compression_ratio": 1.291497975708502 }, { "start": 43.9, "end": 49.78, "text": "就是让他能够实现就是预填充的时候", "chunk": 0, "language": "zh", "avg_logprob": -0.15405842707707332, "no_speech_prob": 2.06851133444097e-11, "compression_ratio": 1.291497975708502 }, { "start": 49.78, "end": 52.28, "text": "它是一个激活参数是8B的", "chunk": 0, "language": "zh", "avg_logprob": -0.15405842707707332, "no_speech_prob": 2.06851133444097e-11, "compression_ratio": 1.291497975708502 }, { "start": 52.28, "end": 56.52, "text": "然后解码的时候它的速度是16B的", "chunk": 0, "language": "zh", "avg_logprob": -0.2427166495353553, "no_speech_prob": 1.675324982908233e-11, "compression_ratio": 1.3779264214046822 }, { "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", "chunk": 0, "language": "zh", "avg_logprob": -0.2427166495353553, "no_speech_prob": 1.675324982908233e-11, "compression_ratio": 1.3779264214046822 }, { "start": 60.2, "end": 62.04, "text": "能够实现大模型的一个推理速度加速", "chunk": 0, "language": "zh", "avg_logprob": -0.2427166495353553, "no_speech_prob": 1.675324982908233e-11, "compression_ratio": 1.3779264214046822 }, { "start": 62.4, "end": 64.76, "text": "然后本次它的这一次的模型", "chunk": 0, "language": "zh", "avg_logprob": -0.2427166495353553, "no_speech_prob": 1.675324982908233e-11, "compression_ratio": 1.3779264214046822 }, { "start": 64.76, "end": 66.24, "text": "已经全量在他们官网上线了", "chunk": 0, "language": "zh", "avg_logprob": -0.2427166495353553, "no_speech_prob": 1.675324982908233e-11, "compression_ratio": 1.3779264214046822 }, { "start": 66.24, "end": 68.04, "text": "而且它官网已经把Deepseek v4 Pro", "chunk": 0, "language": "zh", "avg_logprob": -0.2427166495353553, "no_speech_prob": 1.675324982908233e-11, "compression_ratio": 1.3779264214046822 }, { "start": 68.04, "end": 71.64, "text": "已经路由到它的Deepseek v4 Flash 4.1的模型上面来了", "chunk": 0, "language": "zh", "avg_logprob": -0.2427166495353553, "no_speech_prob": 1.675324982908233e-11, "compression_ratio": 1.3779264214046822 }, { "start": 71.96, "end": 74.36, "text": "而且它经过它的架构优化", "chunk": 0, "language": "zh", "avg_logprob": -0.2427166495353553, "no_speech_prob": 1.675324982908233e-11, "compression_ratio": 1.3779264214046822 }, { "start": 74.44, "end": 76.6, "text": "能够让它的kb缓存降低了", "chunk": 0, "language": "zh", "avg_logprob": -0.2427166495353553, "no_speech_prob": 1.675324982908233e-11, "compression_ratio": 1.3779264214046822 }, { "start": 76.92, "end": 78.64, "text": "就是降低到原来的1%", "chunk": 0, "language": "zh", "avg_logprob": -0.2427166495353553, "no_speech_prob": 1.675324982908233e-11, "compression_ratio": 1.3779264214046822 }, { "start": 78.64, "end": 85.5, "text": "另外一个就是他用了一个压缩的吸收注意力的算法", "chunk": 0, "language": "zh", "avg_logprob": -0.14840677405605798, "no_speech_prob": 1.5683581169878558e-11, "compression_ratio": 1.379591836734694 }, { "start": 85.5, "end": 91.08, "text": "然后能够实现也是一个主要是用来加速的", "chunk": 0, "language": "zh", "avg_logprob": -0.14840677405605798, "no_speech_prob": 1.5683581169878558e-11, "compression_ratio": 1.379591836734694 }, { "start": 91.08, "end": 94.08, "text": "在各个层级用的吸收索引器", "chunk": 0, "language": "zh", "avg_logprob": -0.14840677405605798, "no_speech_prob": 1.5683581169878558e-11, "compression_ratio": 1.379591836734694 }, { "start": 94.08, "end": 96.36, "text": "然后来进行限制索引", "chunk": 0, "language": "zh", "avg_logprob": -0.14840677405605798, "no_speech_prob": 1.5683581169878558e-11, "compression_ratio": 1.379591836734694 }, { "start": 96.36, "end": 98.22, "text": "然后实现的一个推理的一个加速", "chunk": 0, "language": "zh", "avg_logprob": -0.14840677405605798, "no_speech_prob": 1.5683581169878558e-11, "compression_ratio": 1.379591836734694 }, { "start": 98.22, "end": 101.58, "text": "我觉得这块确实在DeepSick在这方面", "chunk": 0, "language": "zh", "avg_logprob": -0.14840677405605798, "no_speech_prob": 1.5683581169878558e-11, "compression_ratio": 1.379591836734694 }, { "start": 101.58, "end": 103.2, "text": "还是做了很多的一个功夫", "chunk": 0, "language": "zh", "avg_logprob": -0.14840677405605798, "no_speech_prob": 1.5683581169878558e-11, "compression_ratio": 1.379591836734694 }, { "start": 103.2, "end": 105.24, "text": "另外的一个就是他的一个MHC", "chunk": 0, "language": "zh", "avg_logprob": -0.14840677405605798, "no_speech_prob": 1.5683581169878558e-11, "compression_ratio": 1.379591836734694 }, { "start": 105.24, "end": 108.32, "text": "这个之前他已经在他的开源的项目里面", "chunk": 0, "language": "zh", "avg_logprob": -0.2320622105752268, "no_speech_prob": 1.950556730023756e-11, "compression_ratio": 1.3431372549019607 }, { "start": 108.32, "end": 109.04, "text": "也多次提到过", "chunk": 0, "language": "zh", "avg_logprob": -0.2320622105752268, "no_speech_prob": 1.950556730023756e-11, "compression_ratio": 1.3431372549019607 }, { "start": 109.04, "end": 112.16, "text": "包括他的enggram的一个存储器", "chunk": 0, "language": "zh", "avg_logprob": -0.2320622105752268, "no_speech_prob": 1.950556730023756e-11, "compression_ratio": 1.3431372549019607 }, { "start": 112.16, "end": 113.16, "text": "然后很早的时候", "chunk": 0, "language": "zh", "avg_logprob": -0.2320622105752268, "no_speech_prob": 1.950556730023756e-11, "compression_ratio": 1.3431372549019607 }, { "start": 113.16, "end": 116.0, "text": "应该是在25年11月份左右的时候", "chunk": 0, "language": "zh", "avg_logprob": -0.2320622105752268, "no_speech_prob": 1.950556730023756e-11, "compression_ratio": 1.3431372549019607 }, { "start": 116.0, "end": 116.92, "text": "他开源出来的", "chunk": 0, "language": "zh", "avg_logprob": -0.2320622105752268, "no_speech_prob": 1.950556730023756e-11, "compression_ratio": 1.3431372549019607 }, { "start": 118.0, "end": 120.36, "text": "另外一个就是他本次是一个全新的", "chunk": 0, "language": "zh", "avg_logprob": -0.2320622105752268, "no_speech_prob": 1.950556730023756e-11, "compression_ratio": 1.3431372549019607 }, { "start": 121.12, "end": 122.76, "text": "完整的多模态的一个大模型", "chunk": 0, "language": "zh", "avg_logprob": -0.2320622105752268, "no_speech_prob": 1.950556730023756e-11, "compression_ratio": 1.3431372549019607 }, { "start": 123.0, "end": 124.52, "text": "他跟原来Deepseek v4 flash", "chunk": 0, "language": "zh", "avg_logprob": -0.2320622105752268, "no_speech_prob": 1.950556730023756e-11, "compression_ratio": 1.3431372549019607 }, { "start": 124.52, "end": 126.52, "text": "相当于在前面加了一个补丁包", "chunk": 0, "language": "zh", "avg_logprob": -0.2320622105752268, "no_speech_prob": 1.950556730023756e-11, "compression_ratio": 1.3431372549019607 }, { "start": 126.52, "end": 128.72, "text": "不一样的是他本次的多模态", "chunk": 0, "language": "zh", "avg_logprob": -0.2320622105752268, "no_speech_prob": 1.950556730023756e-11, "compression_ratio": 1.3431372549019607 }, { "start": 128.84, "end": 132.08, "text": "是用了Deepseek v4的一种技术架构", "chunk": 0, "language": "zh", "avg_logprob": -0.2320622105752268, "no_speech_prob": 1.950556730023756e-11, "compression_ratio": 1.3431372549019607 }, { "start": 132.08, "end": 135.36, "text": "然后它是从预训练的时候", "chunk": 0, "language": "zh", "avg_logprob": -0.29007328090383044, "no_speech_prob": 1.6234015867699902e-11, "compression_ratio": 1.292929292929293 }, { "start": 135.36, "end": 138.28, "text": "就已经开始用到了多谋态的一个预料", "chunk": 0, "language": "zh", "avg_logprob": -0.29007328090383044, "no_speech_prob": 1.6234015867699902e-11, "compression_ratio": 1.292929292929293 }, { "start": 138.28, "end": 139.32, "text": "来进行训练", "chunk": 0, "language": "zh", "avg_logprob": -0.29007328090383044, "no_speech_prob": 1.6234015867699902e-11, "compression_ratio": 1.292929292929293 }, { "start": 139.32, "end": 142.12, "text": "所以他说它是一个原生的多谋态大模型", "chunk": 0, "language": "zh", "avg_logprob": -0.29007328090383044, "no_speech_prob": 1.6234015867699902e-11, "compression_ratio": 1.292929292929293 }, { "start": 142.12, "end": 143.64, "text": "这个是没有问题的", "chunk": 0, "language": "zh", "avg_logprob": -0.29007328090383044, "no_speech_prob": 1.6234015867699902e-11, "compression_ratio": 1.292929292929293 }, { "start": 143.64, "end": 147.08, "text": "我们可以看一下它的一个分数", "chunk": 0, "language": "zh", "avg_logprob": -0.29007328090383044, "no_speech_prob": 1.6234015867699902e-11, "compression_ratio": 1.292929292929293 }, { "start": 147.08, "end": 150.0, "text": "它本次也很克制", "chunk": 0, "language": "zh", "avg_logprob": -0.29007328090383044, "no_speech_prob": 1.6234015867699902e-11, "compression_ratio": 1.292929292929293 }, { "start": 150.0, "end": 153.08, "text": "然后它在Terminal Branch Center上面的话", "chunk": 0, "language": "zh", "avg_logprob": -0.29007328090383044, "no_speech_prob": 1.6234015867699902e-11, "compression_ratio": 1.292929292929293 }, { "start": 153.08, "end": 154.44, "text": "能拿到了一个30分", "chunk": 0, "language": "zh", "avg_logprob": -0.29007328090383044, "no_speech_prob": 1.6234015867699902e-11, "compression_ratio": 1.292929292929293 }, { "start": 154.44, "end": 157.44, "text": "然后比Kimi5.3都要好一些", "chunk": 0, "language": "zh", "avg_logprob": -0.29007328090383044, "no_speech_prob": 1.6234015867699902e-11, "compression_ratio": 1.292929292929293 }, { "start": 157.44, "end": 159.96, "text": "然后比Opus5还是差的很多", "chunk": 0, "language": "zh", "avg_logprob": -0.29007328090383044, "no_speech_prob": 1.6234015867699902e-11, "compression_ratio": 1.292929292929293 }, { "start": 159.96, "end": 160.56, "text": "我们可以看一下", "chunk": 0, "language": "zh", "avg_logprob": -0.29007328090383044, "no_speech_prob": 1.6234015867699902e-11, "compression_ratio": 1.292929292929293 }, { "start": 160.56, "end": 162.9, "text": "然后和GPT5.6", "chunk": 0, "language": "zh", "avg_logprob": -0.34495766005830136, "no_speech_prob": 1.8811646684824268e-11, "compression_ratio": 1.2351190476190477 }, { "start": 162.9, "end": 164.12, "text": "其实GPT6一出来了", "chunk": 0, "language": "zh", "avg_logprob": -0.34495766005830136, "no_speech_prob": 1.8811646684824268e-11, "compression_ratio": 1.2351190476190477 }, { "start": 165.32, "end": 166.64, "text": "所以说还是有很大的一个差距的", "chunk": 0, "language": "zh", "avg_logprob": -0.34495766005830136, "no_speech_prob": 1.8811646684824268e-11, "compression_ratio": 1.2351190476190477 }, { "start": 166.64, "end": 168.24, "text": "在国内在Terminibank上面", "chunk": 0, "language": "zh", "avg_logprob": -0.34495766005830136, "no_speech_prob": 1.8811646684824268e-11, "compression_ratio": 1.2351190476190477 }, { "start": 168.24, "end": 168.96, "text": "当然这些", "chunk": 0, "language": "zh", "avg_logprob": -0.34495766005830136, "no_speech_prob": 1.8811646684824268e-11, "compression_ratio": 1.2351190476190477 }, { "start": 170.08, "end": 170.68, "text": "这叫什么", "chunk": 0, "language": "zh", "avg_logprob": -0.34495766005830136, "no_speech_prob": 1.8811646684824268e-11, "compression_ratio": 1.2351190476190477 }, { "start": 171.32, "end": 174.26, "text": "这些压测的测试的数据集", "chunk": 0, "language": "zh", "avg_logprob": -0.34495766005830136, "no_speech_prob": 1.8811646684824268e-11, "compression_ratio": 1.2351190476190477 }, { "start": 174.26, "end": 176.04, "text": "包括算法都是国外的他们做的", "chunk": 0, "language": "zh", "avg_logprob": -0.34495766005830136, "no_speech_prob": 1.8811646684824268e-11, "compression_ratio": 1.2351190476190477 }, { "start": 176.04, "end": 180.56, "text": "有可能都是由OpenAI或者说Cloud他们定义的", "chunk": 0, "language": "zh", "avg_logprob": -0.34495766005830136, "no_speech_prob": 1.8811646684824268e-11, "compression_ratio": 1.2351190476190477 }, { "start": 180.56, "end": 181.88, "text": "就谁定义谁说算嘛", "chunk": 0, "language": "zh", "avg_logprob": -0.34495766005830136, "no_speech_prob": 1.8811646684824268e-11, "compression_ratio": 1.2351190476190477 }, { "start": 181.88, "end": 183.08, "text": "他肯定的数据会好一些", "chunk": 0, "language": "zh", "avg_logprob": -0.34495766005830136, "no_speech_prob": 1.8811646684824268e-11, "compression_ratio": 1.2351190476190477 }, { "start": 183.76, "end": 185.2, "text": "当然了这个不能绝对", "chunk": 0, "language": "zh", "avg_logprob": -0.34495766005830136, "no_speech_prob": 1.8811646684824268e-11, "compression_ratio": 1.2351190476190477 }, { "start": 185.2, "end": 187.16, "text": "人家的模型确实我们承认他比较先进", "chunk": 0, "language": "zh", "avg_logprob": -0.34495766005830136, "no_speech_prob": 1.8811646684824268e-11, "compression_ratio": 1.2351190476190477 }, { "start": 187.4, "end": 189.88, "text": "另外一个在Deep SWE上面", "chunk": 0, "language": "zh", "avg_logprob": -0.34495766005830136, "no_speech_prob": 1.8811646684824268e-11, "compression_ratio": 1.2351190476190477 }, { "start": 189.88, "end": 191.88, "text": "v1.1的版本测试数据以上的话", "chunk": 0, "language": "zh", "avg_logprob": -0.1997040447435881, "no_speech_prob": 1.795030617202098e-11, "compression_ratio": 1.3231292517006803 }, { "start": 191.88, "end": 193.72, "text": "他应该是拿到了一个最好的成绩", "chunk": 0, "language": "zh", "avg_logprob": -0.1997040447435881, "no_speech_prob": 1.795030617202098e-11, "compression_ratio": 1.3231292517006803 }, { "start": 193.72, "end": 197.84, "text": "然后在安全的测评的上面也拿到了一个88分", "chunk": 0, "language": "zh", "avg_logprob": -0.1997040447435881, "no_speech_prob": 1.795030617202098e-11, "compression_ratio": 1.3231292517006803 }, { "start": 197.84, "end": 201.24, "text": "然后在自动化上面也拿到了一个54分", "chunk": 0, "language": "zh", "avg_logprob": -0.1997040447435881, "no_speech_prob": 1.795030617202098e-11, "compression_ratio": 1.3231292517006803 }, { "start": 201.24, "end": 204.28, "text": "目前我看到网上很多的用户评论说", "chunk": 0, "language": "zh", "avg_logprob": -0.1997040447435881, "no_speech_prob": 1.795030617202098e-11, "compression_ratio": 1.3231292517006803 }, { "start": 204.28, "end": 205.84, "text": "把它接到Deepseek里面去", "chunk": 0, "language": "zh", "avg_logprob": -0.1997040447435881, "no_speech_prob": 1.795030617202098e-11, "compression_ratio": 1.3231292517006803 }, { "start": 205.84, "end": 209.52, "text": "目前接到Codex里面去用Codex Computer Use", "chunk": 0, "language": "zh", "avg_logprob": -0.1997040447435881, "no_speech_prob": 1.795030617202098e-11, "compression_ratio": 1.3231292517006803 }, { "start": 209.52, "end": 213.08, "text": "是目前性价比和成功率最高的模型之一", "chunk": 0, "language": "zh", "avg_logprob": -0.1997040447435881, "no_speech_prob": 1.795030617202098e-11, "compression_ratio": 1.3231292517006803 }, { "start": 213.08, "end": 213.88, "text": "除了Deepseek", "chunk": 0, "language": "zh", "avg_logprob": -0.1997040447435881, "no_speech_prob": 1.795030617202098e-11, "compression_ratio": 1.3231292517006803 }, { "start": 213.88, "end": 217.16, "text": "除了GPT官网之外", "chunk": 0, "language": "zh", "avg_logprob": -0.1997040447435881, "no_speech_prob": 1.795030617202098e-11, "compression_ratio": 1.3231292517006803 }, { "start": 217.16, "end": 221.36, "text": "然后本次算是Deep Seek真正的机身", "chunk": 0, "language": "zh", "avg_logprob": -0.13042109751561928, "no_speech_prob": 1.6376660444406e-11, "compression_ratio": 1.4337349397590362 }, { "start": 221.36, "end": 224.0, "text": "然后世界前沿级的大模型", "chunk": 0, "language": "zh", "avg_logprob": -0.13042109751561928, "no_speech_prob": 1.6376660444406e-11, "compression_ratio": 1.4337349397590362 }, { "start": 224.0, "end": 226.02, "text": "多模态加一照上下文态这个梯堆", "chunk": 0, "language": "zh", "avg_logprob": -0.13042109751561928, "no_speech_prob": 1.6376660444406e-11, "compression_ratio": 1.4337349397590362 }, { "start": 226.02, "end": 228.66, "text": "目前国内相当于是第三家", "chunk": 0, "language": "zh", "avg_logprob": -0.13042109751561928, "no_speech_prob": 1.6376660444406e-11, "compression_ratio": 1.4337349397590362 }, { "start": 228.66, "end": 231.16, "text": "第一家相当于是Kimi", "chunk": 0, "language": "zh", "avg_logprob": -0.13042109751561928, "no_speech_prob": 1.6376660444406e-11, "compression_ratio": 1.4337349397590362 }, { "start": 231.16, "end": 234.02, "text": "第二家相当于是千问", "chunk": 0, "language": "zh", "avg_logprob": -0.13042109751561928, "no_speech_prob": 1.6376660444406e-11, "compression_ratio": 1.4337349397590362 }, { "start": 234.02, "end": 236.4, "text": "但千问38刚开始推出来的不是多模态的", "chunk": 0, "language": "zh", "avg_logprob": -0.13042109751561928, "no_speech_prob": 1.6376660444406e-11, "compression_ratio": 1.4337349397590362 }, { "start": 236.4, "end": 237.0, "text": "我也很奇怪", "chunk": 0, "language": "zh", "avg_logprob": -0.13042109751561928, "no_speech_prob": 1.6376660444406e-11, "compression_ratio": 1.4337349397590362 }, { "start": 237.0, "end": 237.74, "text": "不知道他现在是不是", "chunk": 0, "language": "zh", "avg_logprob": -0.13042109751561928, "no_speech_prob": 1.6376660444406e-11, "compression_ratio": 1.4337349397590362 }, { "start": 237.74, "end": 239.88, "text": "我们就把它放在这一列里面", "chunk": 0, "language": "zh", "avg_logprob": -0.13042109751561928, "no_speech_prob": 1.6376660444406e-11, "compression_ratio": 1.4337349397590362 }, { "start": 239.88, "end": 241.86, "text": "因为我之前说国内只有一家的时候", "chunk": 0, "language": "zh", "avg_logprob": -0.13042109751561928, "no_speech_prob": 1.6376660444406e-11, "compression_ratio": 1.4337349397590362 }, { "start": 241.86, "end": 243.66, "text": "就有人在评论区说千问也是", "chunk": 0, "language": "zh", "avg_logprob": -0.13042109751561928, "no_speech_prob": 1.6376660444406e-11, "compression_ratio": 1.4337349397590362 }, { "start": 243.66, "end": 245.7, "text": "我记得千问38刚出来的时候", "chunk": 0, "language": "zh", "avg_logprob": -0.13042109751561928, "no_speech_prob": 1.6376660444406e-11, "compression_ratio": 1.4337349397590362 }, { "start": 245.7, "end": 246.76, "text": "它是一个文本的多模态", "chunk": 0, "language": "zh", "avg_logprob": -0.13042109751561928, "no_speech_prob": 1.6376660444406e-11, "compression_ratio": 1.4337349397590362 }, { "start": 246.76, "end": 249.38, "text": "因为大家都知道我其实没有做测试", "chunk": 0, "language": "zh", "avg_logprob": -0.17689041991333862, "no_speech_prob": 1.9065514791916094e-11, "compression_ratio": 1.3843537414965987 }, { "start": 249.38, "end": 251.62, "text": "就是没有深入的测试", "chunk": 0, "language": "zh", "avg_logprob": -0.17689041991333862, "no_speech_prob": 1.9065514791916094e-11, "compression_ratio": 1.3843537414965987 }, { "start": 251.62, "end": 252.24, "text": "卖Token Plan", "chunk": 0, "language": "zh", "avg_logprob": -0.17689041991333862, "no_speech_prob": 1.9065514791916094e-11, "compression_ratio": 1.3843537414965987 }, { "start": 252.24, "end": 253.72, "text": "然后限制太多了", "chunk": 0, "language": "zh", "avg_logprob": -0.17689041991333862, "no_speech_prob": 1.9065514791916094e-11, "compression_ratio": 1.3843537414965987 }, { "start": 253.72, "end": 254.58, "text": "所以我就放弃了", "chunk": 0, "language": "zh", "avg_logprob": -0.17689041991333862, "no_speech_prob": 1.9065514791916094e-11, "compression_ratio": 1.3843537414965987 }, { "start": 254.58, "end": 258.64, "text": "然后这个是它的整个的一个测评的一个分数", "chunk": 0, "language": "zh", "avg_logprob": -0.17689041991333862, "no_speech_prob": 1.9065514791916094e-11, "compression_ratio": 1.3843537414965987 }, { "start": 258.64, "end": 260.26, "text": "比如说在各个领域", "chunk": 0, "language": "zh", "avg_logprob": -0.17689041991333862, "no_speech_prob": 1.9065514791916094e-11, "compression_ratio": 1.3843537414965987 }, { "start": 260.26, "end": 261.8, "text": "然后还是非常亮眼的", "chunk": 0, "language": "zh", "avg_logprob": -0.17689041991333862, "no_speech_prob": 1.9065514791916094e-11, "compression_ratio": 1.3843537414965987 }, { "start": 261.8, "end": 262.7, "text": "这个黑色的就表示", "chunk": 0, "language": "zh", "avg_logprob": -0.17689041991333862, "no_speech_prob": 1.9065514791916094e-11, "compression_ratio": 1.3843537414965987 }, { "start": 262.7, "end": 265.5, "text": "它是目前的在已知的测评的这些里面", "chunk": 0, "language": "zh", "avg_logprob": -0.17689041991333862, "no_speech_prob": 1.9065514791916094e-11, "compression_ratio": 1.3843537414965987 }, { "start": 265.5, "end": 269.36, "text": "效果是最好的", "chunk": 0, "language": "zh", "avg_logprob": -0.17689041991333862, "no_speech_prob": 1.9065514791916094e-11, "compression_ratio": 1.3843537414965987 }, { "start": 269.36, "end": 271.58, "text": "我们可以看一下", "chunk": 0, "language": "zh", "avg_logprob": -0.17689041991333862, "no_speech_prob": 1.9065514791916094e-11, "compression_ratio": 1.3843537414965987 }, { "start": 271.58, "end": 273.46, "text": "这个是它的一个base模型", "chunk": 0, "language": "zh", "avg_logprob": -0.17689041991333862, "no_speech_prob": 1.9065514791916094e-11, "compression_ratio": 1.3843537414965987 }, { "start": 273.46, "end": 275.26, "text": "和它的对就是这个模型", "chunk": 0, "language": "zh", "avg_logprob": -0.17689041991333862, "no_speech_prob": 1.9065514791916094e-11, "compression_ratio": 1.3843537414965987 }, { "start": 275.26, "end": 282.92, "text": "它在很多的参数已经非常接近V4 Pro和已经超过的Pro的模型", "chunk": 0, "language": "zh", "avg_logprob": -0.10542356967926025, "no_speech_prob": 2.013647061593904e-11, "compression_ratio": 1.319767441860465 }, { "start": 282.92, "end": 287.06, "text": "所以说他们官网已经把这个Pro路由到这个模型上面来", "chunk": 0, "language": "zh", "avg_logprob": -0.10542356967926025, "no_speech_prob": 2.013647061593904e-11, "compression_ratio": 1.319767441860465 }, { "start": 287.06, "end": 288.9, "text": "第一个是为了降低它的一个推理成本", "chunk": 0, "language": "zh", "avg_logprob": -0.10542356967926025, "no_speech_prob": 2.013647061593904e-11, "compression_ratio": 1.319767441860465 }, { "start": 288.9, "end": 293.0, "text": "第二个也是因为这个模型的能力确实提升非常明显", "chunk": 0, "language": "zh", "avg_logprob": -0.10542356967926025, "no_speech_prob": 2.013647061593904e-11, "compression_ratio": 1.319767441860465 }, { "start": 293.0, "end": 295.8, "text": "还有一个的最主要原因是网上被很多用户吐槽", "chunk": 0, "language": "zh", "avg_logprob": -0.10542356967926025, "no_speech_prob": 2.013647061593904e-11, "compression_ratio": 1.319767441860465 }, { "start": 295.8, "end": 297.4, "text": "Deepstick V4 Pro的幻觉比较严重", "chunk": 0, "language": "zh", "avg_logprob": -0.10542356967926025, "no_speech_prob": 2.013647061593904e-11, "compression_ratio": 1.319767441860465 }, { "start": 297.4, "end": 300.8, "text": "实际上我使用下来整体感觉还是非常不错的", "chunk": 0, "language": "zh", "avg_logprob": -0.10542356967926025, "no_speech_prob": 2.013647061593904e-11, "compression_ratio": 1.319767441860465 }, { "start": 300.8, "end": 303.7, "text": "这些测评的分数我们就不一一讲", "chunk": 0, "language": "zh", "avg_logprob": -0.10542356967926025, "no_speech_prob": 2.013647061593904e-11, "compression_ratio": 1.319767441860465 }, { "start": 303.7, "end": 305.26, "text": "因为刚才我们已经带大家看了一下", "chunk": 0, "language": "zh", "avg_logprob": -0.20697729928152903, "no_speech_prob": 2.389472820996108e-11, "compression_ratio": 1.4099378881987579 }, { "start": 305.26, "end": 307.34, "text": "它跟国外的这些模型的做一些对比", "chunk": 0, "language": "zh", "avg_logprob": -0.20697729928152903, "no_speech_prob": 2.389472820996108e-11, "compression_ratio": 1.4099378881987579 }, { "start": 307.34, "end": 310.06, "text": "其实我们可以看到目前它应该是", "chunk": 0, "language": "zh", "avg_logprob": -0.20697729928152903, "no_speech_prob": 2.389472820996108e-11, "compression_ratio": 1.4099378881987579 }, { "start": 310.06, "end": 313.66, "text": "牺牲到了国内的大模型的第一体队", "chunk": 0, "language": "zh", "avg_logprob": -0.20697729928152903, "no_speech_prob": 2.389472820996108e-11, "compression_ratio": 1.4099378881987579 }, { "start": 313.66, "end": 314.26, "text": "毫无疑问", "chunk": 0, "language": "zh", "avg_logprob": -0.20697729928152903, "no_speech_prob": 2.389472820996108e-11, "compression_ratio": 1.4099378881987579 }, { "start": 314.26, "end": 317.14, "text": "所以说DeepSick还是非常强的", "chunk": 0, "language": "zh", "avg_logprob": -0.20697729928152903, "no_speech_prob": 2.389472820996108e-11, "compression_ratio": 1.4099378881987579 }, { "start": 317.94, "end": 322.66, "text": "目前它官方还没有去推出", "chunk": 0, "language": "zh", "avg_logprob": -0.20697729928152903, "no_speech_prob": 2.389472820996108e-11, "compression_ratio": 1.4099378881987579 }, { "start": 322.66, "end": 324.42, "text": "然后VM和SGLAN部署的", "chunk": 0, "language": "zh", "avg_logprob": -0.20697729928152903, "no_speech_prob": 2.389472820996108e-11, "compression_ratio": 1.4099378881987579 }, { "start": 324.42, "end": 325.66, "text": "然后我相信应该很快", "chunk": 0, "language": "zh", "avg_logprob": -0.20697729928152903, "no_speech_prob": 2.389472820996108e-11, "compression_ratio": 1.4099378881987579 }, { "start": 325.66, "end": 327.66, "text": "VM和SGLAN就有对您的支持", "chunk": 0, "language": "zh", "avg_logprob": -0.20697729928152903, "no_speech_prob": 2.389472820996108e-11, "compression_ratio": 1.4099378881987579 }, { "start": 327.66, "end": 329.34, "text": "就算没有对您也有对一的支持", "chunk": 0, "language": "zh", "avg_logprob": -0.20697729928152903, "no_speech_prob": 2.389472820996108e-11, "compression_ratio": 1.4099378881987579 }, { "start": 329.34, "end": 331.42, "text": "刚才我们在一个我们私有化", "chunk": 0, "language": "zh", "avg_logprob": -0.20697729928152903, "no_speech_prob": 2.389472820996108e-11, "compression_ratio": 1.4099378881987579 }, { "start": 331.42, "end": 333.02, "text": "部署大模型的群里面就有人说", "chunk": 0, "language": "zh", "avg_logprob": -0.20697729928152903, "no_speech_prob": 2.389472820996108e-11, "compression_ratio": 1.4099378881987579 }, { "start": 333.02, "end": 335.38, "text": "今天晚上想蹲一个4.1的本地部署", "chunk": 0, "language": "zh", "avg_logprob": -0.17709701762479893, "no_speech_prob": 1.84413786324944e-11, "compression_ratio": 1.3525835866261398 }, { "start": 335.38, "end": 336.06, "text": "我感觉够呛", "chunk": 0, "language": "zh", "avg_logprob": -0.17709701762479893, "no_speech_prob": 1.84413786324944e-11, "compression_ratio": 1.3525835866261398 }, { "start": 336.06, "end": 337.74, "text": "其实我们手上也有机器", "chunk": 0, "language": "zh", "avg_logprob": -0.17709701762479893, "no_speech_prob": 1.84413786324944e-11, "compression_ratio": 1.3525835866261398 }, { "start": 337.74, "end": 339.9, "text": "但是不打算部署了", "chunk": 0, "language": "zh", "avg_logprob": -0.17709701762479893, "no_speech_prob": 1.84413786324944e-11, "compression_ratio": 1.3525835866261398 }, { "start": 339.9, "end": 342.74, "text": "因为它500多G的话", "chunk": 0, "language": "zh", "avg_logprob": -0.17709701762479893, "no_speech_prob": 1.84413786324944e-11, "compression_ratio": 1.3525835866261398 }, { "start": 342.74, "end": 344.38, "text": "说实话我们就算部署起来的话", "chunk": 0, "language": "zh", "avg_logprob": -0.17709701762479893, "no_speech_prob": 1.84413786324944e-11, "compression_ratio": 1.3525835866261398 }, { "start": 344.38, "end": 346.66, "text": "速度应该也不会很快", "chunk": 0, "language": "zh", "avg_logprob": -0.17709701762479893, "no_speech_prob": 1.84413786324944e-11, "compression_ratio": 1.3525835866261398 }, { "start": 346.66, "end": 348.14, "text": "还有一个原因是", "chunk": 0, "language": "zh", "avg_logprob": -0.17709701762479893, "no_speech_prob": 1.84413786324944e-11, "compression_ratio": 1.3525835866261398 }, { "start": 348.14, "end": 351.78, "text": "它官网的版本应该是加了Dspark", "chunk": 0, "language": "zh", "avg_logprob": -0.17709701762479893, "no_speech_prob": 1.84413786324944e-11, "compression_ratio": 1.3525835866261398 }, { "start": 351.78, "end": 352.54, "text": "然后进行加速", "chunk": 0, "language": "zh", "avg_logprob": -0.17709701762479893, "no_speech_prob": 1.84413786324944e-11, "compression_ratio": 1.3525835866261398 }, { "start": 352.54, "end": 353.82, "text": "但是它开源的版本里面", "chunk": 0, "language": "zh", "avg_logprob": -0.17709701762479893, "no_speech_prob": 1.84413786324944e-11, "compression_ratio": 1.3525835866261398 }, { "start": 353.82, "end": 354.9, "text": "我刚才翻了一下", "chunk": 0, "language": "zh", "avg_logprob": -0.17709701762479893, "no_speech_prob": 1.84413786324944e-11, "compression_ratio": 1.3525835866261398 }, { "start": 354.9, "end": 357.26, "text": "它的文件里面好像没有相关的一些描述", "chunk": 0, "language": "zh", "avg_logprob": -0.17709701762479893, "no_speech_prob": 1.84413786324944e-11, "compression_ratio": 1.3525835866261398 }, { "start": 357.26, "end": 358.9, "text": "我们再等一等看看", "chunk": 0, "language": "zh", "avg_logprob": -0.17709701762479893, "no_speech_prob": 1.84413786324944e-11, "compression_ratio": 1.3525835866261398 }, { "start": 358.9, "end": 361.7, "text": "三方会不会把Dspark给它接入进去", "chunk": 0, "language": "zh", "avg_logprob": -0.17709701762479893, "no_speech_prob": 1.84413786324944e-11, "compression_ratio": 1.3525835866261398 }, { "start": 361.7, "end": 362.94, "text": "他官网的现在这个速度", "chunk": 0, "language": "zh", "avg_logprob": -0.16883450676413142, "no_speech_prob": 2.0662563673945478e-11, "compression_ratio": 1.4631901840490797 }, { "start": 362.94, "end": 365.7, "text": "我相信他应该是把despark已经接入进去了", "chunk": 0, "language": "zh", "avg_logprob": -0.16883450676413142, "no_speech_prob": 2.0662563673945478e-11, "compression_ratio": 1.4631901840490797 }, { "start": 365.7, "end": 367.78, "text": "那好的", "chunk": 0, "language": "zh", "avg_logprob": -0.16883450676413142, "no_speech_prob": 2.0662563673945478e-11, "compression_ratio": 1.4631901840490797 }, { "start": 367.78, "end": 369.86, "text": "如果你对大模型感兴趣的话", "chunk": 0, "language": "zh", "avg_logprob": -0.16883450676413142, "no_speech_prob": 2.0662563673945478e-11, "compression_ratio": 1.4631901840490797 }, { "start": 369.86, "end": 370.62, "text": "可以关注一下", "chunk": 0, "language": "zh", "avg_logprob": -0.16883450676413142, "no_speech_prob": 2.0662563673945478e-11, "compression_ratio": 1.4631901840490797 }, { "start": 370.62, "end": 375.8, "text": "然后来在你的API里面切换一下", "chunk": 0, "language": "zh", "avg_logprob": -0.16883450676413142, "no_speech_prob": 2.0662563673945478e-11, "compression_ratio": 1.4631901840490797 }, { "start": 375.8, "end": 377.76, "text": "其实我上一个视频里面也已经讲到了", "chunk": 0, "language": "zh", "avg_logprob": -0.16883450676413142, "no_speech_prob": 2.0662563673945478e-11, "compression_ratio": 1.4631901840490797 }, { "start": 377.76, "end": 379.12, "text": "就是DeepSick最新的模型", "chunk": 0, "language": "zh", "avg_logprob": -0.16883450676413142, "no_speech_prob": 2.0662563673945478e-11, "compression_ratio": 1.4631901840490797 }, { "start": 379.12, "end": 380.72, "text": "然后我们来也看一下同步的", "chunk": 0, "language": "zh", "avg_logprob": -0.16883450676413142, "no_speech_prob": 2.0662563673945478e-11, "compression_ratio": 1.4631901840490797 }, { "start": 380.72, "end": 383.06, "text": "看一下他们最新的官网的一个描述", "chunk": 0, "language": "zh", "avg_logprob": -0.16883450676413142, "no_speech_prob": 2.0662563673945478e-11, "compression_ratio": 1.4631901840490797 }, { "start": 383.06, "end": 386.88, "text": "我们可以看到他的官网又更新了", "chunk": 0, "language": "zh", "avg_logprob": -0.16883450676413142, "no_speech_prob": 2.0662563673945478e-11, "compression_ratio": 1.4631901840490797 }, { "start": 386.88, "end": 388.16, "text": "原来右侧的是一个招聘的", "chunk": 0, "language": "zh", "avg_logprob": -0.16883450676413142, "no_speech_prob": 2.0662563673945478e-11, "compression_ratio": 1.4631901840490797 }, { "start": 388.16, "end": 388.9, "text": "现在不招了", "chunk": 0, "language": "zh", "avg_logprob": -0.16883450676413142, "no_speech_prob": 2.0662563673945478e-11, "compression_ratio": 1.4631901840490797 }, { "start": 388.9, "end": 389.58, "text": "也不是不招了", "chunk": 0, "language": "zh", "avg_logprob": -0.16883450676413142, "no_speech_prob": 2.0662563673945478e-11, "compression_ratio": 1.4631901840490797 }, { "start": 389.58, "end": 391.06, "text": "他应该是更新了一个", "chunk": 0, "language": "zh", "avg_logprob": -0.16883450676413142, "no_speech_prob": 2.0662563673945478e-11, "compression_ratio": 1.4631901840490797 }, { "start": 391.06, "end": 393.7, "text": "然后我们来到它的开放平台上", "chunk": 0, "language": "zh", "avg_logprob": -0.20984840393066406, "no_speech_prob": 1.490742425336311e-11, "compression_ratio": 1.3205128205128205 }, { "start": 393.7, "end": 394.18, "text": "我们来看一下", "chunk": 0, "language": "zh", "avg_logprob": -0.20984840393066406, "no_speech_prob": 1.490742425336311e-11, "compression_ratio": 1.3205128205128205 }, { "start": 394.18, "end": 396.76, "text": "这个是它的接口文档", "chunk": 0, "language": "zh", "avg_logprob": -0.20984840393066406, "no_speech_prob": 1.490742425336311e-11, "compression_ratio": 1.3205128205128205 }, { "start": 396.76, "end": 398.94, "text": "我们重点看它模型的一个更新", "chunk": 0, "language": "zh", "avg_logprob": -0.20984840393066406, "no_speech_prob": 1.490742425336311e-11, "compression_ratio": 1.3205128205128205 }, { "start": 398.94, "end": 401.56, "text": "目前在它的接口文档里面", "chunk": 0, "language": "zh", "avg_logprob": -0.20984840393066406, "no_speech_prob": 1.490742425336311e-11, "compression_ratio": 1.3205128205128205 }, { "start": 401.56, "end": 402.44, "text": "还是两个模型", "chunk": 0, "language": "zh", "avg_logprob": -0.20984840393066406, "no_speech_prob": 1.490742425336311e-11, "compression_ratio": 1.3205128205128205 }, { "start": 402.44, "end": 405.98, "text": "所以说我们可以看一下模型和价格", "chunk": 0, "language": "zh", "avg_logprob": -0.20984840393066406, "no_speech_prob": 1.490742425336311e-11, "compression_ratio": 1.3205128205128205 }, { "start": 405.98, "end": 409.36, "text": "冰干压我记得还有一个", "chunk": 0, "language": "zh", "avg_logprob": -0.20984840393066406, "no_speech_prob": 1.490742425336311e-11, "compression_ratio": 1.3205128205128205 }, { "start": 409.36, "end": 412.98, "text": "Dipsick Flash", "chunk": 0, "language": "zh", "avg_logprob": -0.20984840393066406, "no_speech_prob": 1.490742425336311e-11, "compression_ratio": 1.3205128205128205 }, { "start": 412.98, "end": 416.36, "text": "Dipsick V4 Pro", "chunk": 0, "language": "zh", "avg_logprob": -0.20984840393066406, "no_speech_prob": 1.490742425336311e-11, "compression_ratio": 1.3205128205128205 }, { "start": 416.36, "end": 419.0, "text": "它把它的模型", "chunk": 0, "language": "zh", "avg_logprob": -0.20984840393066406, "no_speech_prob": 1.490742425336311e-11, "compression_ratio": 1.3205128205128205 }, { "start": 419.0, "end": 419.94, "text": "这个4.1 Flash", "chunk": 0, "language": "zh", "avg_logprob": -0.20984840393066406, "no_speech_prob": 1.490742425336311e-11, "compression_ratio": 1.3205128205128205 }, { "start": 419.94, "end": 421.78, "text": "已经直接切换成这个了", "chunk": 0, "language": "zh", "avg_logprob": -0.1639974989542147, "no_speech_prob": 3.1539919981282694e-11, "compression_ratio": 1.5578947368421052 }, { "start": 421.78, "end": 423.46, "text": "我们来看一下它的", "chunk": 0, "language": "zh", "avg_logprob": -0.1639974989542147, "no_speech_prob": 3.1539919981282694e-11, "compression_ratio": 1.5578947368421052 }, { "start": 423.46, "end": 428.18, "text": "模型名称使用Deepseek V4 Flash", "chunk": 0, "language": "zh", "avg_logprob": -0.1639974989542147, "no_speech_prob": 3.1539919981282694e-11, "compression_ratio": 1.5578947368421052 }, { "start": 428.18, "end": 429.82, "text": "就可以接入到新的模型里面来了", "chunk": 0, "language": "zh", "avg_logprob": -0.1639974989542147, "no_speech_prob": 3.1539919981282694e-11, "compression_ratio": 1.5578947368421052 }, { "start": 429.82, "end": 433.18, "text": "然后旧的V4 Flash和Flash Vision XP", "chunk": 0, "language": "zh", "avg_logprob": -0.1639974989542147, "no_speech_prob": 3.1539919981282694e-11, "compression_ratio": 1.5578947368421052 }, { "start": 433.18, "end": 434.22, "text": "还可以调用", "chunk": 0, "language": "zh", "avg_logprob": -0.1639974989542147, "no_speech_prob": 3.1539919981282694e-11, "compression_ratio": 1.5578947368421052 }, { "start": 434.22, "end": 435.78, "text": "它应该是已经把它路由到", "chunk": 0, "language": "zh", "avg_logprob": -0.1639974989542147, "no_speech_prob": 3.1539919981282694e-11, "compression_ratio": 1.5578947368421052 }, { "start": 435.78, "end": 437.82, "text": "Deepseek V4 E Flash这个模型上面来", "chunk": 0, "language": "zh", "avg_logprob": -0.1639974989542147, "no_speech_prob": 3.1539919981282694e-11, "compression_ratio": 1.5578947368421052 }, { "start": 437.82, "end": 439.74, "text": "然后它的统一名称叫Deepseek V4 Flash", "chunk": 0, "language": "zh", "avg_logprob": -0.1639974989542147, "no_speech_prob": 3.1539919981282694e-11, "compression_ratio": 1.5578947368421052 }, { "start": 439.74, "end": 442.2, "text": "没有V4就是Deepseek Flash模型", "chunk": 0, "language": "zh", "avg_logprob": -0.1639974989542147, "no_speech_prob": 3.1539919981282694e-11, "compression_ratio": 1.5578947368421052 }, { "start": 442.2, "end": 443.46, "text": "新的模型的话", "chunk": 0, "language": "zh", "avg_logprob": -0.1639974989542147, "no_speech_prob": 3.1539919981282694e-11, "compression_ratio": 1.5578947368421052 }, { "start": 443.46, "end": 445.24, "text": "你可以直接使用Deepseek V4 Flash", "chunk": 0, "language": "zh", "avg_logprob": -0.1639974989542147, "no_speech_prob": 3.1539919981282694e-11, "compression_ratio": 1.5578947368421052 }, { "start": 445.24, "end": 446.72, "text": "也可以使用Deepseek Flash", "chunk": 0, "language": "zh", "avg_logprob": -0.1639974989542147, "no_speech_prob": 3.1539919981282694e-11, "compression_ratio": 1.5578947368421052 }, { "start": 446.72, "end": 447.94, "text": "都可以来做我介入", "chunk": 0, "language": "zh", "avg_logprob": -0.1639974989542147, "no_speech_prob": 3.1539919981282694e-11, "compression_ratio": 1.5578947368421052 }, { "start": 447.94, "end": 450.46, "text": "我再给大家带大家来看一下", "chunk": 0, "language": "zh", "avg_logprob": -0.26285418024602925, "no_speech_prob": 9.779906919027415e-12, "compression_ratio": 1.1363636363636365 }, { "start": 450.46, "end": 453.24, "text": "我们来到我们的7166里面", "chunk": 0, "language": "zh", "avg_logprob": -0.26285418024602925, "no_speech_prob": 9.779906919027415e-12, "compression_ratio": 1.1363636363636365 }, { "start": 453.24, "end": 455.78, "text": "带大家来做一次配置一下", "chunk": 0, "language": "zh", "avg_logprob": -0.26285418024602925, "no_speech_prob": 9.779906919027415e-12, "compression_ratio": 1.1363636363636365 }, { "start": 455.78, "end": 458.18, "text": "其他的所有的三方的工具都是一样的", "chunk": 0, "language": "zh", "avg_logprob": -0.26285418024602925, "no_speech_prob": 9.779906919027415e-12, "compression_ratio": 1.1363636363636365 }, { "start": 458.18, "end": 461.26, "text": "OK", "chunk": 0, "language": "zh", "avg_logprob": -0.26285418024602925, "no_speech_prob": 9.779906919027415e-12, "compression_ratio": 1.1363636363636365 }, { "start": 462.563, "end": 468.123, "text": "OK。我們在這裡面添加一下模型。OK。添加模型。來做一下檢測。", "chunk": 1, "language": "zh", "avg_logprob": -0.37530784084372326, "no_speech_prob": 8.726156949800945e-12, "compression_ratio": 1.3885714285714286 }, { "start": 468.123, "end": 473.403, "text": "连接失败", "chunk": 1, "language": "zh", "avg_logprob": -0.12156238775143678, "no_speech_prob": 1.6825181872737183e-11, "compression_ratio": 1.3605015673981191 }, { "start": 473.403, "end": 476.503, "text": "我这个账户应该是欠费的", "chunk": 1, "language": "zh", "avg_logprob": -0.12156238775143678, "no_speech_prob": 1.6825181872737183e-11, "compression_ratio": 1.3605015673981191 }, { "start": 476.503, "end": 479.703, "text": "正好我今天也开玩了一个", "chunk": 1, "language": "zh", "avg_logprob": -0.12156238775143678, "no_speech_prob": 1.6825181872737183e-11, "compression_ratio": 1.3605015673981191 }, { "start": 479.703, "end": 484.903, "text": "我自己的一个本地的大模型的一个推理的框架", "chunk": 1, "language": "zh", "avg_logprob": -0.12156238775143678, "no_speech_prob": 1.6825181872737183e-11, "compression_ratio": 1.3605015673981191 }, { "start": 484.903, "end": 486.143, "text": "然后同时也支持云端", "chunk": 1, "language": "zh", "avg_logprob": -0.12156238775143678, "no_speech_prob": 1.6825181872737183e-11, "compression_ratio": 1.3605015673981191 }, { "start": 486.143, "end": 487.203, "text": "先简单给大家介绍一下", "chunk": 1, "language": "zh", "avg_logprob": -0.12156238775143678, "no_speech_prob": 1.6825181872737183e-11, "compression_ratio": 1.3605015673981191 }, { "start": 487.203, "end": 488.623, "text": "我们这里面有对话", "chunk": 1, "language": "zh", "avg_logprob": -0.12156238775143678, "no_speech_prob": 1.6825181872737183e-11, "compression_ratio": 1.3605015673981191 }, { "start": 488.623, "end": 490.383, "text": "有agent还有视频的通话的", "chunk": 1, "language": "zh", "avg_logprob": -0.12156238775143678, "no_speech_prob": 1.6825181872737183e-11, "compression_ratio": 1.3605015673981191 }, { "start": 490.383, "end": 492.423, "text": "包括这种可以本地和云端的模式", "chunk": 1, "language": "zh", "avg_logprob": -0.12156238775143678, "no_speech_prob": 1.6825181872737183e-11, "compression_ratio": 1.3605015673981191 }, { "start": 492.423, "end": 493.563, "text": "可以跟他先打电话", "chunk": 1, "language": "zh", "avg_logprob": -0.12156238775143678, "no_speech_prob": 1.6825181872737183e-11, "compression_ratio": 1.3605015673981191 }, { "start": 493.563, "end": 494.743, "text": "但是我这个电脑没在我旁边", "chunk": 1, "language": "zh", "avg_logprob": -0.12156238775143678, "no_speech_prob": 1.6825181872737183e-11, "compression_ratio": 1.3605015673981191 }, { "start": 494.743, "end": 495.903, "text": "所以说没办法进行沟通", "chunk": 1, "language": "zh", "avg_logprob": -0.12156238775143678, "no_speech_prob": 1.6825181872737183e-11, "compression_ratio": 1.3605015673981191 }, { "start": 495.903, "end": 496.643, "text": "我们先把它挂了", "chunk": 1, "language": "zh", "avg_logprob": -0.12156238775143678, "no_speech_prob": 1.6825181872737183e-11, "compression_ratio": 1.3605015673981191 }, { "start": 496.643, "end": 497.863, "text": "回头给单独给大家测试", "chunk": 1, "language": "zh", "avg_logprob": -0.12156238775143678, "no_speech_prob": 1.6825181872737183e-11, "compression_ratio": 1.3605015673981191 }, { "start": 497.863, "end": 501.423, "text": "然后这一块是生成本地去生成TTS", "chunk": 1, "language": "zh", "avg_logprob": -0.11268265581361338, "no_speech_prob": 2.3915595198653605e-11, "compression_ratio": 1.632398753894081 }, { "start": 501.423, "end": 502.863, "text": "然后可以使用Edge TTS", "chunk": 1, "language": "zh", "avg_logprob": -0.11268265581361338, "no_speech_prob": 2.3915595198653605e-11, "compression_ratio": 1.632398753894081 }, { "start": 502.863, "end": 504.903, "text": "也可以使用所谓OpenAI兼容的协议的", "chunk": 1, "language": "zh", "avg_logprob": -0.11268265581361338, "no_speech_prob": 2.3915595198653605e-11, "compression_ratio": 1.632398753894081 }, { "start": 504.903, "end": 506.343, "text": "来进行本地去生成音频", "chunk": 1, "language": "zh", "avg_logprob": -0.11268265581361338, "no_speech_prob": 2.3915595198653605e-11, "compression_ratio": 1.632398753894081 }, { "start": 506.343, "end": 508.043, "text": "这块是生图的", "chunk": 1, "language": "zh", "avg_logprob": -0.11268265581361338, "no_speech_prob": 2.3915595198653605e-11, "compression_ratio": 1.632398753894081 }, { "start": 508.043, "end": 510.503, "text": "然后也可以用本地的MRX生图", "chunk": 1, "language": "zh", "avg_logprob": -0.11268265581361338, "no_speech_prob": 2.3915595198653605e-11, "compression_ratio": 1.632398753894081 }, { "start": 510.503, "end": 511.823, "text": "也可以使用云端的一个生图", "chunk": 1, "language": "zh", "avg_logprob": -0.11268265581361338, "no_speech_prob": 2.3915595198653605e-11, "compression_ratio": 1.632398753894081 }, { "start": 511.823, "end": 514.663, "text": "然后我们可以看一下之前历史生图的一些效果", "chunk": 1, "language": "zh", "avg_logprob": -0.11268265581361338, "no_speech_prob": 2.3915595198653605e-11, "compression_ratio": 1.632398753894081 }, { "start": 514.663, "end": 515.543, "text": "这是生图的", "chunk": 1, "language": "zh", "avg_logprob": -0.11268265581361338, "no_speech_prob": 2.3915595198653605e-11, "compression_ratio": 1.632398753894081 }, { "start": 515.543, "end": 516.483, "text": "还有一个OCR", "chunk": 1, "language": "zh", "avg_logprob": -0.11268265581361338, "no_speech_prob": 2.3915595198653605e-11, "compression_ratio": 1.632398753894081 }, { "start": 516.483, "end": 517.703, "text": "然后上传文档", "chunk": 1, "language": "zh", "avg_logprob": -0.11268265581361338, "no_speech_prob": 2.3915595198653605e-11, "compression_ratio": 1.632398753894081 }, { "start": 517.703, "end": 518.883, "text": "然后进行OCR提取", "chunk": 1, "language": "zh", "avg_logprob": -0.11268265581361338, "no_speech_prob": 2.3915595198653605e-11, "compression_ratio": 1.632398753894081 }, { "start": 518.883, "end": 519.523, "text": "这是这个", "chunk": 1, "language": "zh", "avg_logprob": -0.11268265581361338, "no_speech_prob": 2.3915595198653605e-11, "compression_ratio": 1.632398753894081 }, { "start": 519.523, "end": 520.483, "text": "除了这个之外", "chunk": 1, "language": "zh", "avg_logprob": -0.11268265581361338, "no_speech_prob": 2.3915595198653605e-11, "compression_ratio": 1.632398753894081 }, { "start": 520.483, "end": 521.483, "text": "我们还有翻译", "chunk": 1, "language": "zh", "avg_logprob": -0.11268265581361338, "no_speech_prob": 2.3915595198653605e-11, "compression_ratio": 1.632398753894081 }, { "start": 521.483, "end": 522.763, "text": "这是翻译的内容", "chunk": 1, "language": "zh", "avg_logprob": -0.11268265581361338, "no_speech_prob": 2.3915595198653605e-11, "compression_ratio": 1.632398753894081 }, { "start": 522.763, "end": 523.523, "text": "中英文翻译", "chunk": 1, "language": "zh", "avg_logprob": -0.11268265581361338, "no_speech_prob": 2.3915595198653605e-11, "compression_ratio": 1.632398753894081 }, { "start": 523.523, "end": 525.103, "text": "还有一个提示词", "chunk": 1, "language": "zh", "avg_logprob": -0.11268265581361338, "no_speech_prob": 2.3915595198653605e-11, "compression_ratio": 1.632398753894081 }, { "start": 525.103, "end": 526.403, "text": "我们有一个提示词仓库", "chunk": 1, "language": "zh", "avg_logprob": -0.11268265581361338, "no_speech_prob": 2.3915595198653605e-11, "compression_ratio": 1.632398753894081 }, { "start": 526.403, "end": 527.263, "text": "然后把它集成进来了", "chunk": 1, "language": "zh", "avg_logprob": -0.11268265581361338, "no_speech_prob": 2.3915595198653605e-11, "compression_ratio": 1.632398753894081 }, { "start": 527.263, "end": 532.383, "text": "而这个仓库也已经在GitHub已经开完了", "chunk": 1, "language": "zh", "avg_logprob": -0.2568565845489502, "no_speech_prob": 1.411099882775435e-11, "compression_ratio": 1.5246636771300448 }, { "start": 532.383, "end": 535.423, "text": "我们在这个里面就是我们的一个模型", "chunk": 1, "language": "zh", "avg_logprob": -0.2568565845489502, "no_speech_prob": 1.411099882775435e-11, "compression_ratio": 1.5246636771300448 }, { "start": 535.423, "end": 536.023, "text": "我们可以看一下", "chunk": 1, "language": "zh", "avg_logprob": -0.2568565845489502, "no_speech_prob": 1.411099882775435e-11, "compression_ratio": 1.5246636771300448 }, { "start": 536.023, "end": 538.263, "text": "Deepseek v4 flash 还有Deepseek flash", "chunk": 1, "language": "zh", "avg_logprob": -0.2568565845489502, "no_speech_prob": 1.411099882775435e-11, "compression_ratio": 1.5246636771300448 }, { "start": 538.263, "end": 539.863, "text": "这个模型都已经可以了", "chunk": 1, "language": "zh", "avg_logprob": -0.2568565845489502, "no_speech_prob": 1.411099882775435e-11, "compression_ratio": 1.5246636771300448 }, { "start": 539.863, "end": 541.863, "text": "在我们这个工具里面也可以调用了", "chunk": 1, "language": "zh", "avg_logprob": -0.2568565845489502, "no_speech_prob": 1.411099882775435e-11, "compression_ratio": 1.5246636771300448 }, { "start": 541.863, "end": 542.703, "text": "我们来给大家测一下", "chunk": 1, "language": "zh", "avg_logprob": -0.2568565845489502, "no_speech_prob": 1.411099882775435e-11, "compression_ratio": 1.5246636771300448 }, { "start": 542.703, "end": 545.383, "text": "来切到Deepseek v4 flash", "chunk": 1, "language": "zh", "avg_logprob": -0.2568565845489502, "no_speech_prob": 1.411099882775435e-11, "compression_ratio": 1.5246636771300448 }, { "start": 545.383, "end": 547.023, "text": "Deepseek flash 这个模型", "chunk": 1, "language": "zh", "avg_logprob": -0.2568565845489502, "no_speech_prob": 1.411099882775435e-11, "compression_ratio": 1.5246636771300448 }, { "start": 547.023, "end": 548.303, "text": "OK 我们可以看一下", "chunk": 1, "language": "zh", "avg_logprob": -0.2568565845489502, "no_speech_prob": 1.411099882775435e-11, "compression_ratio": 1.5246636771300448 }, { "start": 548.303, "end": 554.383, "text": "不欠", "chunk": 1, "language": "zh", "avg_logprob": -0.2568565845489502, "no_speech_prob": 1.411099882775435e-11, "compression_ratio": 1.5246636771300448 }, { "start": 554.383, "end": 558.583, "text": "是这个吧", "chunk": 1, "language": "zh", "avg_logprob": -0.37461244607273536, "no_speech_prob": 1.0614504067363306e-11, "compression_ratio": 1.3132530120481927 }, { "start": 566.583, "end": 568.383, "text": "我这个好像有点小问题", "chunk": 1, "language": "zh", "avg_logprob": -0.37461244607273536, "no_speech_prob": 1.0614504067363306e-11, "compression_ratio": 1.3132530120481927 }, { "start": 568.383, "end": 569.743, "text": "我再试一下我们的agent", "chunk": 1, "language": "zh", "avg_logprob": -0.37461244607273536, "no_speech_prob": 1.0614504067363306e-11, "compression_ratio": 1.3132530120481927 }, { "start": 572.803, "end": 573.803, "text": "应该是可以调通的", "chunk": 1, "language": "zh", "avg_logprob": -0.37461244607273536, "no_speech_prob": 1.0614504067363306e-11, "compression_ratio": 1.3132530120481927 }, { "start": 574.623, "end": 575.363, "text": "是没有问题的", "chunk": 1, "language": "zh", "avg_logprob": -0.37461244607273536, "no_speech_prob": 1.0614504067363306e-11, "compression_ratio": 1.3132530120481927 }, { "start": 575.363, "end": 576.923, "text": "然后这个对话应该也是可以的", "chunk": 1, "language": "zh", "avg_logprob": -0.37461244607273536, "no_speech_prob": 1.0614504067363306e-11, "compression_ratio": 1.3132530120481927 }, { "start": 576.923, "end": 578.063, "text": "我再新建一个对话试一下", "chunk": 1, "language": "zh", "avg_logprob": -0.37461244607273536, "no_speech_prob": 1.0614504067363306e-11, "compression_ratio": 1.3132530120481927 }, { "start": 578.063, "end": 578.603, "text": "稍等一下", "chunk": 1, "language": "zh", "avg_logprob": -0.37461244607273536, "no_speech_prob": 1.0614504067363306e-11, "compression_ratio": 1.3132530120481927 }, { "start": 579.783, "end": 580.843, "text": "我们来切换一下", "chunk": 1, "language": "zh", "avg_logprob": -0.37461244607273536, "no_speech_prob": 1.0614504067363306e-11, "compression_ratio": 1.3132530120481927 }, { "start": 580.843, "end": 585.323, "text": "ok 是没有问题的", "chunk": 1, "language": "zh", "avg_logprob": -0.2853579158671418, "no_speech_prob": 2.6970170888263034e-11, "compression_ratio": 1.2700296735905046 }, { "start": 585.323, "end": 588.723, "text": "那好的 今天就介绍在这", "chunk": 1, "language": "zh", "avg_logprob": -0.2853579158671418, "no_speech_prob": 2.6970170888263034e-11, "compression_ratio": 1.2700296735905046 }, { "start": 588.723, "end": 589.203, "text": "谢谢大家", "chunk": 1, "language": "zh", "avg_logprob": -0.2853579158671418, "no_speech_prob": 2.6970170888263034e-11, "compression_ratio": 1.2700296735905046 }, { "start": 589.203, "end": 590.163, "text": "如果大家感兴趣的话", "chunk": 1, "language": "zh", "avg_logprob": -0.2853579158671418, "no_speech_prob": 2.6970170888263034e-11, "compression_ratio": 1.2700296735905046 }, { "start": 590.163, "end": 591.243, "text": "我们这里面的支持的", "chunk": 1, "language": "zh", "avg_logprob": -0.2853579158671418, "no_speech_prob": 2.6970170888263034e-11, "compression_ratio": 1.2700296735905046 }, { "start": 591.243, "end": 592.323, "text": "我们自己的自带的", "chunk": 1, "language": "zh", "avg_logprob": -0.2853579158671418, "no_speech_prob": 2.6970170888263034e-11, "compression_ratio": 1.2700296735905046 }, { "start": 592.323, "end": 593.883, "text": "还Deepseek 千问 质朴", "chunk": 1, "language": "zh", "avg_logprob": -0.2853579158671418, "no_speech_prob": 2.6970170888263034e-11, "compression_ratio": 1.2700296735905046 }, { "start": 593.883, "end": 596.043, "text": "Kimi 豆包 文心一言 腾讯", "chunk": 1, "language": "zh", "avg_logprob": -0.2853579158671418, "no_speech_prob": 2.6970170888263034e-11, "compression_ratio": 1.2700296735905046 }, { "start": 596.043, "end": 598.723, "text": "Binimax 科大讯飞和借阅的", "chunk": 1, "language": "zh", "avg_logprob": -0.2853579158671418, "no_speech_prob": 2.6970170888263034e-11, "compression_ratio": 1.2700296735905046 }, { "start": 598.723, "end": 601.123, "text": "所以这些厂商的大模型在支持", "chunk": 1, "language": "zh", "avg_logprob": -0.2853579158671418, "no_speech_prob": 2.6970170888263034e-11, "compression_ratio": 1.2700296735905046 }, { "start": 601.123, "end": 602.443, "text": "我们内置了一个agent", "chunk": 1, "language": "zh", "avg_logprob": -0.2853579158671418, "no_speech_prob": 2.6970170888263034e-11, "compression_ratio": 1.2700296735905046 }, { "start": 602.443, "end": 603.603, "text": "这个agent是py agent", "chunk": 1, "language": "zh", "avg_logprob": -0.2853579158671418, "no_speech_prob": 2.6970170888263034e-11, "compression_ratio": 1.2700296735905046 }, { "start": 604.643, "end": 606.363, "text": "然后还有一些其他的相关的一些功能", "chunk": 1, "language": "zh", "avg_logprob": -0.2853579158671418, "no_speech_prob": 2.6970170888263034e-11, "compression_ratio": 1.2700296735905046 }, { "start": 606.363, "end": 607.963, "text": "然后主打的就是本地推理", "chunk": 1, "language": "zh", "avg_logprob": -0.2853579158671418, "no_speech_prob": 2.6970170888263034e-11, "compression_ratio": 1.2700296735905046 }, { "start": 607.963, "end": 610.363, "text": "那再顺便再多说两句吧", "chunk": 1, "language": "zh", "avg_logprob": -0.2853579158671418, "no_speech_prob": 2.6970170888263034e-11, "compression_ratio": 1.2700296735905046 }, { "start": 610.363, "end": 612.203, "text": "我们这里面在本地区部署大模型", "chunk": 1, "language": "zh", "avg_logprob": -0.2533349841052949, "no_speech_prob": 2.9050383626438858e-11, "compression_ratio": 1.4294117647058824 }, { "start": 612.203, "end": 614.123, "text": "支持Nameter CPP VRM和SGLAN", "chunk": 1, "language": "zh", "avg_logprob": -0.2533349841052949, "no_speech_prob": 2.9050383626438858e-11, "compression_ratio": 1.4294117647058824 }, { "start": 614.123, "end": 615.083, "text": "然后进行部署", "chunk": 1, "language": "zh", "avg_logprob": -0.2533349841052949, "no_speech_prob": 2.9050383626438858e-11, "compression_ratio": 1.4294117647058824 }, { "start": 615.083, "end": 616.283, "text": "然后在模型库里面", "chunk": 1, "language": "zh", "avg_logprob": -0.2533349841052949, "no_speech_prob": 2.9050383626438858e-11, "compression_ratio": 1.4294117647058824 }, { "start": 616.283, "end": 618.683, "text": "我们有一些自己挑选的模型", "chunk": 1, "language": "zh", "avg_logprob": -0.2533349841052949, "no_speech_prob": 2.9050383626438858e-11, "compression_ratio": 1.4294117647058824 }, { "start": 618.683, "end": 619.963, "text": "可以在这里面来进行管理", "chunk": 1, "language": "zh", "avg_logprob": -0.2533349841052949, "no_speech_prob": 2.9050383626438858e-11, "compression_ratio": 1.4294117647058824 }, { "start": 619.963, "end": 621.323, "text": "然后也可以在这里面进行搜索", "chunk": 1, "language": "zh", "avg_logprob": -0.2533349841052949, "no_speech_prob": 2.9050383626438858e-11, "compression_ratio": 1.4294117647058824 }, { "start": 621.323, "end": 622.523, "text": "下载GGOF格式的", "chunk": 1, "language": "zh", "avg_logprob": -0.2533349841052949, "no_speech_prob": 2.9050383626438858e-11, "compression_ratio": 1.4294117647058824 }, { "start": 623.083, "end": 626.603, "text": "然后这个模型也可以在国内进行下载", "chunk": 1, "language": "zh", "avg_logprob": -0.2533349841052949, "no_speech_prob": 2.9050383626438858e-11, "compression_ratio": 1.4294117647058824 }, { "start": 626.603, "end": 627.643, "text": "我已经做到加速", "chunk": 1, "language": "zh", "avg_logprob": -0.2533349841052949, "no_speech_prob": 2.9050383626438858e-11, "compression_ratio": 1.4294117647058824 }, { "start": 627.643, "end": 629.403, "text": "然后我们把本地的这些模型", "chunk": 1, "language": "zh", "avg_logprob": -0.2533349841052949, "no_speech_prob": 2.9050383626438858e-11, "compression_ratio": 1.4294117647058824 }, { "start": 629.403, "end": 629.963, "text": "启动起来之后", "chunk": 1, "language": "zh", "avg_logprob": -0.2533349841052949, "no_speech_prob": 2.9050383626438858e-11, "compression_ratio": 1.4294117647058824 }, { "start": 629.963, "end": 631.323, "text": "我们做了一个统一的网关", "chunk": 1, "language": "zh", "avg_logprob": -0.2533349841052949, "no_speech_prob": 2.9050383626438858e-11, "compression_ratio": 1.4294117647058824 }, { "start": 631.323, "end": 633.563, "text": "然后接着OpenAPI的一个协议的", "chunk": 1, "language": "zh", "avg_logprob": -0.2533349841052949, "no_speech_prob": 2.9050383626438858e-11, "compression_ratio": 1.4294117647058824 }, { "start": 633.563, "end": 634.803, "text": "VModel然后Chatter", "chunk": 1, "language": "zh", "avg_logprob": -0.2533349841052949, "no_speech_prob": 2.9050383626438858e-11, "compression_ratio": 1.4294117647058824 }, { "start": 634.803, "end": 637.483, "text": "然后Response和VMessage和Speech", "chunk": 1, "language": "zh", "avg_logprob": -0.2533349841052949, "no_speech_prob": 2.9050383626438858e-11, "compression_ratio": 1.4294117647058824 }, { "start": 637.483, "end": 641.043, "text": "然后转翻译的和生图的都有了", "chunk": 1, "language": "zh", "avg_logprob": -0.22055775511498546, "no_speech_prob": 9.320535662715734e-12, "compression_ratio": 0.9185185185185185 }, { "start": 641.043, "end": 644.023, "text": "后面我们还会打算把minimax H3接进来", "chunk": 1, "language": "zh", "avg_logprob": -0.22055775511498546, "no_speech_prob": 9.320535662715734e-12, "compression_ratio": 0.9185185185185185 }, { "start": 644.023, "end": 644.543, "text": "好的", "chunk": 1, "language": "zh", "avg_logprob": -0.22055775511498546, "no_speech_prob": 9.320535662715734e-12, "compression_ratio": 0.9185185185185185 }, { "start": 644.543, "end": 645.623, "text": "今天就介绍在这", "chunk": 1, "language": "zh", "avg_logprob": -0.22055775511498546, "no_speech_prob": 9.320535662715734e-12, "compression_ratio": 0.9185185185185185 }, { "start": 645.623, "end": 646.003, "text": "谢谢大家", "chunk": 1, "language": "zh", "avg_logprob": -0.22055775511498546, "no_speech_prob": 9.320535662715734e-12, "compression_ratio": 0.9185185185185185 } ], "segmented_transcription": { "status": "SEGMENTED_TRANSCRIPT_OK", "created_at": "2026-09-11T08:42:56", "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_hint": "zh", "condition_on_previous_text": false, "chunk_count": 2, "resumed_chunks": 0, "merged_segment_count": 276, "cleaned_segment_count": 276, "removed_hallucination_segment_count": 0, "final_segment_end": 646.003, "global_repeat_run": 1, "global_repeat_text": "各位朋友大家好。歡迎大家關注昆蓬道。今天告訴大家一個好消息。Deepseek V4.1 Flash這個模型正式開源了。", "tail_repeat_run": 1, "tail_repeat_text": "还Deepseek 千问 质朴", "hallucination_region_count": 0, "hallucination_regions": [], "qc_ok": true, "qc_errors": [] } }