Event ·Multi-source

Meta发布新研究：字节模型蒸馏后，性能突破GPT-4

Meta发布新研究，将字节跳动开发的GLM-4.5V模型进行蒸馏，新模型在MMLU基准测试中取得82.1分，超越GPT-4的81.5分。该研究展示了模型蒸馏技术如何提升模型性能。

Current read

Meta搞了个新研究，把字节跳动的GLM-4.5V模型蒸馏了一下，结果性能超过了GPT-4，挺有意思的。

7 sources58° AI heatLast updated 2026/9/15 06:37:16

Evidence

7 sources

Primary source量子位

2026/9/15 06:37:16View source

Related source 1elvis

2026/9/14 15:14:02View source

Related source 2arXiv: DeepSeek

2026/9/14 18:58:09View source

Related source 3IT之家

2026/9/15 05:36:29View source

Related source 4Liputan6 Tekno

2026/9/16 03:00:28View source

Related source 5AI Will

2026/9/16 05:41:06View source

Related source 6小互

2026/9/17 02:10:49View source

Conflict check

Dedup data only shows that these sources discuss the same event, not that they agree. There is no structured supporting or refuting evidence yet, so no conflict is inferred automatically.

Open the story