0
Applied AI·August 7, 2026·1 min read

Sources: ByteDance is pretraining an AI model with up to 10T parameters, roughly 3x larger than Kimi K3 and larger than the 8T estimate for Anthropic's Mythos 5

Share

A 10T-parameter pretraining run from ByteDance is another data point that Chinese consumer platforms are willing to spend at frontier-model scale, not just fine-tune. For Western operators, assume TikTok’s parent will be a first-class model vendor and content engine, not just a distribution channel.

Applied AI

Sources: Alibaba plans to ask heavy commercial users of its next Qwen open model for a share of revenue; Moonshot's Kimi K3 requires up to a 30% revenue share

“Open” foundation models are converging on usage- or revenue-based tolls for serious commercial deployment — Alibaba reportedly targeting major Qwen users while Moonshot’s Kimi K3 already takes up to 30%. If your product P&L assumes free or one-time-cost open models, you need to re-underwrite unit economics and vendor leverage now.