AI
search ESC
person
logout
المدونة chevron_left أخبار chevron_left A 2.8-Trillion-Parameter Open Model Just Beat …
folder أخبار

A 2.8-Trillion-Parameter Open Model Just Beat Claude and GPT in a Blind Test

person
Learn-with-us
calendar_today 24 Jul 2026 schedule 1 د قراءة visibility 51 مشاهدة
A 2.8-Trillion-Parameter Open Model Just Beat Claude and GPT in a Blind Test

On July 16, developers ranked code submissions at the Frontend Code Arena without knowing which model wrote them. The entry that beat both Claude Fable 5 and GPT-5.6 Sol — two of the most expensive models you can rent — was Kimi K3, an open-weight model from Moonshot AI in Beijing.

At 2.8 trillion parameters, it's the largest open-source model ever released, and the first to cross the 3-trillion-parameter class. Full weights land publicly on July 27 — meaning the whole community will start compressing and optimizing it almost immediately, the same way it happened with Llama and DeepSeek.

The real story isn't the size. It's that the gap between open and closed frontier AI, once measured in years, is now down to months. That shift matters more for the field than any single benchmark win.

star قيّم المقال Rate this article

سجّل الدخول لتقييم المقال

تسجيل الدخول

share شارك المقال

chat_bubble التعليقات (0)

سجّل الدخول لإضافة تعليق

تسجيل الدخول

لا توجد تعليقات بعد. كن أول من يعلّق!