Kimi K3, Qwen 3.8, and Anthropic's (Potential) Unravelling
Points and comments are a snapshot, not live.
New open models Kimi K3 and Qwen 3.8 challenge Anthropic's position as a top-tier AI vendor.
Moonshot Labs' Kimi K3 and Alibaba's Qwen 3.8, both close to Anthropic's Fable 5 in performance, will have public weights released. This threatens model-only vendors like Anthropic, which faces high inference costs and lacks infrastructure ownership. Anthropic's strategy relies on regulation or AGI, while OpenAI diversifies into hardware and data centers. The article argues that owning infrastructure provides a cost advantage and that model-only providers risk being squeezed by open models and vertical-specific competitors.
What commenters are saying
Commenters split on Anthropic's vulnerability. Some argue user loyalty and the value of frontier models for complex tasks justify premium pricing, citing personal workflows where $200/month is easily offset. Others counter that open models like Kimi K3 will enable cheap alternatives, that Anthropic's model costs are unsustainable for many users, and that Chinese labs can pivot to consumer hardware. A key point: the heavy subsidization of current pricing means real costs could be far higher, and open models will soon be served at near cost.