Moonshot AI, a Beijing-based startup, has unveiled its Kimi K3 AI model, which is causing ripples across the tech and financial industries. The K3 model is substantial, boasting approximately 2.8 trillion parameters, significantly more than its Chinese competitors like DeepSeek V4 Pro (1.6 trillion) and Zhipu AI's GLM-5 series (740 billion). It features native multimodal capabilities, allowing it to process both images and text simultaneously, and an ultra-large context window handling up to 1 million tokens at once. Moonshot AI claims Kimi K3 is the world's first "3-trillion-parameter-class open-weight AI model" and is specialized in long-form coding, complex reasoning, and expert knowledge tasks.
The Kimi K3's performance is approaching that of top-tier US models. Artificial Analysis, an AI performance evaluator, rated Kimi K3's intelligence index at 57 points. While slightly below Anthropic's Claude Fable 5 (60 points), OpenAI's GPT-5.6 Sol (59 points), and GPT-5.6 Sol X-High (58 points), it surpassed xAI's Grok 4.5 and major models from Meta and Google. The model's "front-end coding capability" topped Arena's rankings, indicating its strong performance in this area. Though Moonshot AI acknowledges Kimi K3 still trails the most capable proprietary models overall, its competitive performance has taken the US tech industry by surprise.
The release of Kimi K3 immediately impacted US markets, with the Nasdaq index falling 1.40%, the S&P 500 falling 1.01%, and the Dow Jones Industrial Average falling 0.77% on the 17th. Major chip stocks also saw declines, with Nvidia dropping 2.2%, Applied Materials 5.6%, Intel 2%, and SanDisk 4%. This sell-off reflects growing concerns among investors that the rapid advancement of near-free open AI models could weaken the profitability of subscription-based closed models from companies like OpenAI, Anthropic, and Google. Additionally, there are worries that the timeline for recovering significant AI infrastructure investments might be delayed.
Despite being a Chinese AI model, Kimi K3's pricing strategy deviates from typical Chinese offerings. Its API fees are around $3 per 1 million input tokens and $15 per 1 million output tokens, which is similar to OpenAI's GPT-5.6 Sol but more expensive than other Chinese open models. This suggests Moonshot AI is targeting the enterprise AI market rather than engaging in simple price competition. The use of a mixture-of-experts structure, activating 16 of its 896 specialist components for each task, contributes to its 2.5 times scaling efficiency compared to its previous Kimi K2 model. The company's refusal to disclose the chips used for Kimi K3, amid US regulations on high-performance AI chip exports, has led to speculation that Chinese-made AI semiconductors, such as Huawei Ascend, and Chinese HBM may play a significant role, which could pose a challenge for Samsung Electronics and SK hynix.
