English
← Back to timeline

first half of 2026

Open-Weight Models Publish at Frontier Scale

Era: The LLM Era · 2020s–Present

Through the first half of 2026, open-weight releases from DeepSeek, Alibaba Qwen, Moonshot, Zhipu, and Meta were published at a scale previously seen only in closed systems, with weights and model cards downloadable rather than described. Whether any of them match the closed frontier depends on which benchmark index is used and how it is weighted, so this entry records what was released and under what licence, not a ranking.

DeepSeek / Meta / Alibaba

Status
Developing
Content type
interpretation
Sources reviewed

Sources and support

  1. DeepSeek-R1 paper

    DeepSeek / arXiv · Source type: paper

    Reports evaluations for a prominent open-weight reasoning model.

    Accessed:

  2. Kimi-K3 model card

    Moonshot AI via Hugging Face · Source type: primary-announcement

    Downloadable weights and a model card documenting parameter count, context window, and licence.

    Published: ·Accessed: