开启时会模糊预览图，关闭后正常显示
github：https://t.co/iTk9XtdRUZ HF：https://t.co/Bb9CPV6WfF
致力于让每个想拥抱AI的人都能找到适合自己的AI产品，助力企业定制AIGC应用
美团刚刚开源了一款低延迟全模态大模型：LongCat-Flash-Omni，128K上下文+支持超8分钟实时音视频交互 能听能看能说，毫秒级延迟，可以拿来搭建实时语音助手、直播弹幕机器人、音视频质检等 在OmniBench上超Gemini-2.5-Flash、Qwen3-Omni，接近Gemini 2.5 Pro；在MMBench与Gemini、GPT-4o、Qwen-VL相当；MVBench上超Gemini 2.5 Pro和 GPT-4o 通用知识接近DeepSeek V3.1、Qwen3等 总参数560B，激活27B，基于其LongCat-Flash的ScMoE 架构 #大模型 #LongCatFlashOmni
github：https://t.co/iTk9XtdRUZ HF：https://t.co/Bb9CPV6WfF
I can feel it now — being able to quiet your mind and focus is, in itself, a kind of mastery.
Growth Coach｜Helping creators build their personal brand on X 增长教练｜帮助创作者在X上打造个人品牌 公众号：PandaTalk8 X 增长群群主
Pro Russia account be like
Research Scientist @meta (FAIR), Prof. @Unige_en, co-founder @nc_shape. I like reality.
Fast Models optimize for Time to First Token Fast Agents optimize for Time to Tell The Agent It Is Wrong 2 fast wrongs often do make a fast right
achieve ambition with intentionality, intensity, & integrity - @dxtipshq - @sveltesociety - @aidotengineer - @latentspacepod - @cognition + @smol_ai
Super excited that @karpathy noticed our work! Hopefully it helps the broader community realize that *precision* deserves a place in our design space.
Research engineer, RL believer @SeaAIL | 💻 📄 🏸
