返回时间轴
2025-04-01

Meta releases Llama 4, the first natively multimodal open-weight model family

Capability Breakthrough

事件摘要

In April 2025, Meta released Llama 4, the first open-weight model family built with natively multimodal architecture — processing text, images, and video jointly from the ground up rather than adding vision as a separate module. The family included Llama 4 Maverick (flagship reasoning) and Llama 4 Scout (efficient deployment). Building on the Llama 3.1 405B breakthrough, Llama 4 established open-weight multimodal AI as the new baseline for the open-source ecosystem.

影响评估

  • Capability Leap +1 · Medium-term

    First natively multimodal open-weight model family. End-to-end multimodal training achieved significantly better visual reasoning than patched approaches. Established multimodality as default for open-weight ecosystem.

    Affected Groups: AI developers, open-source community, researchers

共识度与来源

重要度 L1
分类 Capability Breakthrough
共识度 Broad Consensus
影响指数 5/10
  • 1
    The Llama 4 herd — Meta AI Blog

    URL: https://ai.meta.com/blog/llama-4-multimodal-intelligence/

    The Llama 4 herd: The beginning of a new era of natively multimodal AI innovation.
    Reference Evidence Citation logged Live source

关联事件

← Origins
2023-02-24 L2

Meta releases LLaMA, triggering the open-weight LLM revolution

Meta published LLaMA (Large Language Model Meta AI) in February 2023, a set of foundation language models from 7B to 65B parameters trained on publicly available data only. LLaMA-13B outperformed GPT-3 (175B) on most benchmarks, proving that smaller, well-trained models could rival much larger ones. Although originally released under a non-commercial research license, the weights were leaked and quickly fine-tuned by the community — spawning the open-weight LLM ecosystem that reshaped the AI industry.

2024-07-23 L2

Meta releases Llama 3.1 405B, the first frontier-level open-source AI model

On July 23, 2024, Meta released Llama 3.1 405B — the first open-weight model to match closed-source frontier models like GPT-4 and Claude 3.5 Sonnet on key benchmarks. With 405 billion dense parameters and a 128K token context window, it proved that open-source AI could compete at the highest level. The release reshaped the AI industry by demonstrating that open models could match proprietary ones, accelerating the open-weight ecosystem that would define 2025.

Local Lineage Meta releases Llama ... 2025-04-01 · L1 Meta releases LLaM... 2023-02-24 · L2 Meta releases Llam... 2024-07-23 · L2