China's Moonshot unveils world's largest open AI model, closing in on US rivals

• Kimi K3 features 2.8 trillion parameters and a 1 million-token context window, ranking first in web interface benchmarks and second overall behind only Fable 5 in third-party evaluations
• The release follows Anthropic's Fable and Mythos models being withdrawn by the U.S. government due to security concerns, and comes as multiple Chinese firms (MiniMax, Z.ai) accelerate releases of trillion-parameter systems
• Moonshot, backed by Alibaba and Tencent, is reportedly seeking $2 billion in funding at a $30 billion valuation ahead of a potential Hong Kong IPO

Market Summary: China's Moonshot Unveils World's Largest Open AI Model

Key Development:

Chinese AI startup Moonshot launched Kimi K3 on July 17, 2026, a 2.8 trillion-parameter model described as the world's largest open-weight AI system. The company claims performance approaching U.S. rival Anthropic's Fable model, signaling China's narrowing technological gap with American AI leaders.

Technical Specifications:

• 2.8 trillion parameters, approaching the 3 trillion threshold
• 1 million-token context window for enhanced information processing
• Designed for advanced reasoning, long-horizon coding, and knowledge work
• Outperformed OpenAI's Opus 4.8, GPT 5.6 Sol, and GPT 5.5 in GPU optimization tests

Market Context:

The release follows Anthropic's Fable and Mythos models being withdrawn by the U.S. government last month due to security concerns. Chinese firms including Moonshot, Z.ai, and MiniMax are releasing increasingly powerful models at lower costs, challenging Western assumptions that Chinese developers lag American peers by months.

Third-Party Performance:

Independent benchmarks confirm strong capabilities: Arena.ai ranked Kimi K3 first in web interface-building; Vals AI placed it second overall behind Fable 5; Artificial Analysis found performance comparable to GPT-5.5 and Claude Opus 4.8.

Company Financials:

Backed by Alibaba and Tencent, Moonshot is reportedly seeking $2 billion in funding at a $30 billion valuation ahead of a potential Hong Kong listing.

Competitive Landscape:

MiniMax is developing a 2.7 trillion-parameter model for Q3 2026 release. Previous Chinese leaders Meituan's LongCat-2.0 and DeepSeek's V4-Pro featured 1.6 trillion parameters. The shift toward trillion-parameter systems reflects demand for autonomous systems handling complex reasoning tasks.