Beijing-based Moonshot AI unveiled a model that benchmarks say rivals GPT-5.6 and Claude Fable 5, rattling global AI and semiconductor markets.
China just dropped another model that is making investors nervous. Moonshot AI, a Beijing-based startup founded in 2023, unveiled its Kimi K3 model in mid-July 2026, and the market reaction was swift and unambiguous: rival AI stocks fell sharply, semiconductor names followed, and the broader conversation about AI valuations got a lot more uncomfortable.
The numbers behind the model are hard to ignore. Kimi K3 features 2.8 trillion parameters and a 1 million token context window, using a Mixture-of-Experts architecture with open weights. On benchmarks spanning coding and long-horizon knowledge tasks, the model reportedly matches or outperforms leading US models including GPT-5.6 and Claude Fable 5.
What happened in the market #
The immediate fallout landed hardest on Moonshot’s domestic competitors. Z.ai saw its shares drop approximately 27% following the announcement. MiniMax fell roughly 16%.
Moonshot’s pricing adds another layer of disruption. Kimi K3 is priced at around $3 per million input tokens and $15 per million output tokens, positioning it as a cost-competitive alternative to models that currently command premium pricing.
Who is Moonshot AI and why it matters now #
Moonshot AI launched in 2023 and raised $1 billion from Alibaba in 2024, pushing its valuation to $2.5 billion at the time. Subsequent investment rounds drove its valuation to a range reported between $18 billion and $31.5 billion.
The full open-weight release of Kimi K3 is expected by July 27, 2026. Open weights allow developers to run the model themselves, fine-tune it, and deploy it without paying API fees.
What investors should watch #
Look at the open-weight release date first. If the full model weights arrive July 27 as anticipated and independent evaluations confirm the benchmark claims, the conversation about AI stock valuations will almost certainly intensify.
For companies building AI-powered products, Kimi K3 is potentially good news on costs. Locking infrastructure to a single US provider looks less defensible if open-weight alternatives deliver comparable results at lower inference costs. Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our