home.social

#minimaxm1 — Public Fediverse posts

Live and recent posts from across the Fediverse tagged #minimaxm1, aggregated by home.social.

fetched live
  1. 個人主觀感覺排名: #Qwen3 > #DeepSeekV31 ~ #Glm45 > #MiniMaxM1 ~ #KimiK2 ~ #Ernie45

    實際上它們之間的差距很小,如果 GPT-5 是90分它們基本上都有70分左右,Qwen3 有80分左右,#DeepSeekV31 用得還不多大概也是70分左右,沒有很強。 #DeepSeekV31最強是它的超高性價比,一如以往R1推出的時候,一口氣把價格壓下來。

  2. #MiniMaxM1 用過也不差,個人覺得和 #Glm45 是差不多的水平,一樣是比不上 #DeepSeek

  3. MiniMax-M1:闪电注意力重塑大模型推理效率,百万上下文时代来临,附技术报告英中对照版 一、核心创新:闪电注意力 + 混合架构 1. 闪电注意力(Light...

    #LLm #大模型 #预训练模型 #MiniMax #MiniMax-M1

    Origin | Interest | Match
  4. MiniMax-M1: Разбираем архитектуру, ломающую законы масштабирования (и наш VRAM)

    В мире LLM доминирует квадратичная сложность, ограничивающая контекст. Но MiniMax-M1 бросает вызов: миллион токенов, низкие затраты. Разбираем гибридную архитектуру с Lightning Attention, новый алгоритм CISPO и инженерные прорывы, делающие эту модель уникальной.

    habr.com/ru/articles/923588/

    #minimaxm1 #LLM_архитектура #Lightning_Attention #mixtureofexperts #масштабирование_LLM

  5. Can China’s MiniMax-M1 AI Topple US Rivals? We Put It to the Test In brief MiniMax-M1 excels at coding and agent tasks, but creative writers will want to look elsewhere. Despite marketing claims,...

    #NFTs #China's #MiniMaxM1 #put #rivals #test #Topple

    Origin | Interest | Match
  6. 🥳🤖 Behold, the MiniMax-M1: yet another gloriously named Frankenstein of jargon that promises to solve all your coding woes while draining your soul one AI-generated line at a time. Because clearly, what the world needed was an "open-weight, large-scale hybrid-attention reasoning model" that's harder to understand than quantum physics—and twice as useful. 🚀💻
    github.com/MiniMax-AI/MiniMax- #MiniMaxM1 #AIcoding #TechHumor #HybridAttention #QuantumPhysics #HackerNews #ngated

Share on Mastodon

Enter the server where you have an account.