Moonshot's Kimi K3 set to rival Anthropic's Opus 4.8

Moonshot AI, a Chinese AI laboratory, is preparing to release Kimi K3, an open-weight model that Financial Times reports is expected to match or exceed the performance of Anthropic's Opus 4.8. The model will contain between 2 trillion and 3 trillion parameters, positioning it as the largest open-weight AI model developed in China. Release is scheduled for the coming days. The move follows strong market reception of Moonshot's earlier Kimi K2 models, which have ranked highly on industry benchmarks and demonstrated capabilities approaching those of leading frontier models. Moonshot is simultaneously raising fresh capital, with the new funding round valuing the company at $31.5 billion（約5兆円）—a significant jump from the $20 billion（約3.2兆円） valuation the company achieved in May when it raised $2 billion（約3200億円）.

The timing of Kimi K3's release reflects broader industry anxieties about the dominance of expensive, closed-source AI systems. Executives and industry leaders increasingly fear that AI labs like OpenAI and Anthropic may extract proprietary data that clients submit for use with products such as ChatGPT and Claude. This concern has prompted a renewed focus on open-source alternatives. Industry leaders are now pitching their own open-source products as safer alternatives or recommending that companies deploy cheaper open-source models—such as those developed by DeepSeek, Z.ai, or Moonshot—and train them internally for their specific needs. Kimi K3 positions Moonshot as a credible contender in this shift away from commercial closed-source providers.

Moonshot AI's Kimi K3 represents a significant escalation in the open-source AI competition against large US-based labs. The company's Kimi K2 models have already gained traction in benchmarks and demonstrated competitive capabilities, but Kimi K3 is positioned as a step closer to closing the gap with leading closed-source models. This comes at a moment when enterprise confidence in commercial AI providers is being tested: executives and industry leaders are increasingly concerned about potential data extraction by OpenAI, Anthropic, and similar labs, prompting a shift toward open-source alternatives that companies can deploy and fine-tune independently. The timing of Moonshot's funding at $31.5 billion（約5兆円）—a 57.5% increase from its May round—signals strong investor belief in this market shift. Open-source alternatives like those from DeepSeek and Moonshot are emerging as viable, cost-effective options for organizations willing to shoulder the responsibility of training and deploying models themselves, threatening the commercial moat that closed-source labs have relied upon.