Analysis: The gap between open-source AI and cutting-edge models has shrunk from 12 months to 4 months. By the end of the year, top-tier intelligence may be available for free download.

GLM0%
GPT0%

BlockBeats News, June 21st, Crypto and tech analyst Miles Deutscher cited Epoch AI's Capability Index data, indicating that the gap in capabilities between open-source AI models and closed-source cutting-edge large models is rapidly narrowing—this gap was about 12 months two years ago, and has now compressed to about 4 months.

From the development trajectory, open-source models have caught up from Llama 2-70B to GLM 5.2, Kimi K2.6, etc., while closed-source frontiers have evolved to the scale of GPT-5.5 Pro. At the current catching-up rate, Deutscher predicts that in about another 6 months, top-tier intelligence equivalent to Claude Fable level will be freely available for public download in open-source form.

This trend implies that the capability advantage once seen as a core moat by closed-source giants is rapidly diminishing, and the rise of the open-source ecosystem may profoundly reshape the AI business landscape and computing power competition.

---------------------------------
Click the original text link below to join the Actionable Insights Beating · Feishu AI News Channel, monitoring global AI hotspots and news 24/7.

來源：BlockBeats

免責聲明：當前內容均來自第三方觀點或由AI直接翻譯第三方觀點，CoinEx不保證內容的真實性、準確性和原創性，不構成CoinEx相關的任何投資建議。數字資產價格波動劇烈，請注意潛在風險。

Founder of Brainance, Tang Jie: After the open-source release of GLM-5.2, the performance gap with OpenAI and Anthropic will gradually narrow.

Intellectual Chief Scientist Responds to Musk's "China Big Model Q1 Catching Up to Fable" Remark: It Won't Take That Long

熱搜榜

幣種
價格
24H漲跌