whatmodel.sh: In October 2026, the performance gap between Chinese and US AI large models narrowed to 1 month

Add new crowd-testing promotional copy Add new crowd-testing promotional copy end Add App-exclusive traffic redirect

This chart originally compared the gap between open-source AI and closed-source AI. Earlier, American AI giants were also champions of open source, but after 2023 most American AI shifted to closed source. The American companies still doing open source are not only few in number, but their performance is not outstanding either. At present, it is Chinese companies that are carrying the banner of open-source AI.

The latest large model compared in this chart is still DeepSeek V4.1 Flash-1017, released not long ago. If the Kimi K3.1 Pro released in the early hours of this morning is taken into account, the performance gap will narrow a bit further, though the time gap is still about 1 month.

Looking back at 2026, domestic AI faced considerable pressure from American AI this year, even greater than in previous years, especially in April when Anthropic released the Mythos large model; at that time many people were quite despairing about domestic AI.

After the DeepSeek V4 series was released at the end of April, a US congressional report once listed it as the most advanced domestic large model at the time, but it was still 8 months behind the United States.

In the second half of the year, with a series of offensives including the release of the Mythes/Fable large models, Opus 5.0-5.5, and GPT-5.6 through GPT-6, domestic AI went through several months of silence, especially since July when Kimi released K3.

Recently, with two consecutive releases of DeepSeek V4.1 Flash and the release of Kimi K3.1 Pro, the situation has finally begun to turn around. Next there are also Zhipu's GLM-5.5, Alibaba's Qwen4 series, and DeepSeek V4.1 Pro; the performance ceiling of domestic AI can rise even higher, and GLM-5.5 may even truly reach the top spot.

The AI race is not a 100-meter sprint but a marathon lasting many years. Even if domestic AI catches up with Mythes/Fable this year, the competitive pressure afterward will still not be easy. The two OA companies possess computing power resources and financing far beyond those of domestic companies, and they come with a global market passport, so their revenue and profits are therefore also far higher than those of domestic companies.

But these hardware gaps will not interrupt the progress of domestic AI. The previous US plan was to ensure at least a 2-year lead over Chinese AI through chip blockades and semiconductor technology sanctions, and to lead China by more than 10 years in semiconductor technology.

Today these two goals have become impossible to achieve, and as time passes, domestic self-developed technology is breaking through faster and faster. These suppression sanctions in the US plan are becoming increasingly ineffective, and the day when domestic AI and even semiconductor technology catch up and surpass is not far off. Statement: This is an exclusive article from Sina.com; unauthorized reproduction is prohibited. tech_web_article_v2015_block_bottom Non-targeted 300*250 button 17/09 wenjing begin

Non-targeted 300*250 button end