China’s Kimi K3 Challenges Proprietary AI Leaders
Moonshot AI is preparing to release its advanced Kimi K3 model as open-weight software, according to reports from Bloomberg. The expected release date for the model’s weights is Monday. The Kimi K3 has garnered significant attention since its debut on July 16 due to its ability to combine high, near-frontier artificial intelligence performance with reduced operational costs. Developers have specifically commended the model for its robust capabilities in reasoning, coding, and agentic tasks, which underscores the rapid progress within China’s domestic AI sector.
The decision to make Kimi K3 open-source allows developers worldwide the freedom to modify and download the model. This move is anticipated to accelerate adoption rates and significantly increase competitive pressure on the proprietary offerings provided by leading American AI corporations, including OpenAI and Anthropic.
Geopolitical and Policy Reactions to Open AI
The growing success of open-weight AI models originating from China is drawing attention from policymakers and industry figures in the United States, prompting a debate over potential regulatory restrictions. Separately, U.S. Treasury Secretary Scott Bessent stated last week that the Trump administration plans to investigate allegations of American intellectual property theft by Chinese companies. He added that the government possesses the authority to sanction these entities due to such theft.
In response to these potential restrictions, a coalition of over 20 companies, which includes Microsoft and Nvidia, issued a joint letter. This letter strongly advised policymakers to avoid imposing “premature restrictions” on open-weight artificial intelligence models, arguing that such actions could potentially “stifle competition or drive innovation overseas.” Both Nvidia CEO Jensen Huang and Microsoft CEO Satya Nadella publicly shared this letter on their personal social media accounts, while Elon Musk, founder of Tesla and SpaceX, also voiced his full support for the open model-weight strategy on X.
The Race Among US AI Giants
The major American AI firms are also advancing their technology rapidly. Anthropic launched Claude Opus 5 on July 21, followed by OpenAI’s release of GPT-5.6 on July 24. Meanwhile, Google has not yet launched its anticipated Gemini 3.5 Pro model. Reports emerged last week indicating a delay, as DeepMind continues to improve the model’s coding performance. Consequently, Google is currently utilizing updated, less expensive Gemini Flash models while its competitors advance their most sophisticated AI systems.
Financial Developments and Market Sentiment
Investors are closely monitoring the progress of OpenAI and Anthropic, particularly as both companies approach potential public listings. Both organizations had confidentially submitted IPO documentation to regulators last month. However, reports have surfaced suggesting that OpenAI might postpone its listing until the following year, deviating from the initially anticipated fourth-quarter timeline.
In terms of valuation, Anthropic currently holds an edge over OpenAI. According to data provided by Nasdaq Private Market, Anthropic was valued at $1.12 trillion, compared to OpenAI’s private market valuation of $847.96 billion. Furthermore, Anthropic reported in April that its annualized revenue run rate had tripled to $30 billion, surpassing OpenAI’s approximate $24 billion in Annual Recurring Revenue (ARR).
Retail sentiment tracked on Stocktwits showed a ‘neutral’ rating for OpenAI, a ‘bullish’ rating for Anthropic, and an ‘extremely bullish’ rating for GOOGL.
