Llama 4
AI模型
多模态
上一代
Meta's MoE generation — and the stumble that triggered the superintelligence pivot.
您正在查看一条 AI 模型记录：谁构建了它、它能做什么，以及其背后的硬件与数据。
Llama 4 (April 2025) moved to mixture-of-experts with native multimodality: Scout (109B/17B active, 10M-token context claim) and Maverick (400B/17B active), with the 2-trillion-parameter Behemoth teacher announced but never released. A mixed reception — benchmark controversy, delayed flagship — preceded Zuckerberg's Superintelligence Labs reorganization and hiring spree. In April 2026 MSL replaced the Llama line with Muse Spark — launched, in a historic reversal, without open weights.
它如何与整个技术栈相连
本记录的关联链接一览——点击分组可直接跳转。
开发方
供电
训练于
前代产品
来源
记录审阅截至 2026-08-31 — 查看 方法论.