Moonshot reportedly trained Kimi K3 on Nvidia Blackwell chips for its next model
RebeccaBellan · x · 2026-07-28
Moonshot reportedly trained Kimi K3 on Nvidia chips, including Blackwell
A report highlighted in the post says Moonshot trained Kimi K3 on Nvidia chips, including Blackwell parts, despite export restrictions. The article cites several people familiar with the matter.
The screenshots add more detail: Kimi K3 is described as the world’s largest open-source model at 2.8 trillion parameters, and Moonshot is reportedly seeking more Blackwell access to prepare K4. Another screenshot says the model was trained partly in China and partly at a Thailand data center, reflecting the difficulty of moving pre-training data overseas. The same report also notes that Alibaba’s Qwen3.8-Max was trained on Nvidia chips, including Blackwell.
Related event: Moonshot Reportedly Used Nvidia Blackwell for Kimi K3(2 posts)→
More from Infra
- Neocloud Stocks Slashed Again: CRWV -10%, NBIS -9%, IREN -8% in One Day — JOBhakdi · 2026-10-16
- Neocloud stocks sink on Gartner's "80% will fail" report; analyst calls it garbage — McDonaghMatthew · 2026-10-16
- JapanFold Claims One Tenstorrent Server Beats an Nvidia DGX B200 on Bio Models at a Third of the Price — DavidBennett__ · 2026-10-16
- vLLM ships Day-0 RL support for Beam in SkyRL, +62% LoRA decode throughput — vllm_project · 2026-10-16
- DFlash2 speeds Beam decode up to 3.03× across benchmarks, vLLM reports — vllm_project · 2026-10-16
- Inside Beam: 501B params, only 23B active per token via 8-of-256 expert routing — vllm_project · 2026-10-16