Kimi K3 launch and rapid uptake
Chinese startup Moonshot AI released its Kimi K3 model this week, and the company suspended new subscriptions within two days after demand overwhelmed its computing capacity, CNN Business reported on July 23, 2026. The model is drawing attention for reportedly matching or, in some cases, outperforming the latest iterations of OpenAI's GPT and Anthropic's Claude, with access offered for free. Kimi K3 has 2.8 trillion parameters and uses a mixture-of-experts architecture in which 16 of 896 specialist components activate per request, while native vision and a one-million-token context window support larger workloads, according to a July 24, 2026 industry report. Moonshot set launch pricing at $0.30 per million cache-hit input tokens, $3 per million cache-miss input tokens, and $15 per million output tokens, and plans to release the full trained weights on July 27 so the model can be downloaded and self-hosted.
Microsoft evaluates Kimi K3 for selected Copilot tasks
Microsoft is weighing whether some Copilot requests could be routed from OpenAI and Anthropic models to Kimi K3, according to a July 24, 2026 report citing The Information, though no deployment decision has been confirmed. The report cites an unconfirmed estimate that a partial move could reduce annual AI inference costs by up to $600 million, though the eligible workloads and underlying assumptions are not disclosed. Microsoft said paid Microsoft 365 Copilot seats exceeded 20 million in fiscal Q3 2026, a base over which small per-answer cost differences could become financially material. Kimi K3 ranked second on Vals AI's index, third on the Artificial Analysis Intelligence Index, and first in Frontend Code Arena in those third-party benchmarks. Microsoft's existing Azure model router can already restrict routing to a chosen set of models, but any production use of Kimi K3 would have to clear quality, latency, safety, governance, and capacity tests before handling Copilot traffic.
White House weighs restrictions on Chinese AI
The release is intensifying pressure on the Trump administration to decide whether to regulate or rely on Chinese AI, techxplore.com reported on July 24, 2026. White House Office of Science and Technology Policy director Michael Kratsios accused Moonshot of illicitly distilling Anthropic's most powerful model and of circumventing US export curbs on advanced Nvidia AI chips. The Commerce Department's Bureau of Industry and Security has formally opened an investigation into Chinese firms including Moonshot over their use of those chips, a spokesperson told The Information. Treasury Secretary Scott Bessent has threatened sanctions against China, and the administration is also weighing a ban or curbs on foreign-made open-source models that are often built through distillation, according to the same reporting.
Anthropic allegations and prior warnings
Earlier in 2026, Anthropic sent letters to US lawmakers formally accusing Moonshot, DeepSeek, and MiniMax of industrial-scale distillation of its Claude models, according to techxplore.com. Anthropic's head of public policy, Sarah Heck, described illicit distillation as "IP theft and industrial espionage" and "a national challenge that creates serious national security risks for the United States and democratic allies" in posts on X. The underlying practice is widely used across the AI industry, but US officials argue Chinese firms have deployed it at scale to copy proprietary American systems, framing it as a core justification for tighter export controls and chip-sale restrictions on China.
Industry debate over open-source access
A coalition of 179 startups organized through the Little Tech Association wrote to the Trump administration urging it to forgo any outright ban or strict curbs on foreign-made models, arguing that "denying American startups access to models available abroad would stifle competition, entrench incumbents and function as a tax on intelligence," techxplore.com reported on July 24, 2026. Former White House AI policy chief David Sacks wrote on X that "the Kimi panic needs to stop" and that President Trump's light-touch approach is working. Nvidia CEO Jensen Huang, whose company supplies chips to both Chinese and US AI developers, told Axios that American companies should "absolutely" be allowed to use Chinese AI models because they are excellent. The debate is unfolding as OpenAI and Anthropic prepare for expected public listings, with Wall Street scrutinizing their business performance more closely than before their IPOs.
Share this article
