July 20, 2026
Open-model mayhem
Kimi K3, Qwen 3.8, and Anthropic's (Potential) Unravelling
Cheap new AI stars just crashed the party—and commenters are split on who's doomed
TLDR: Kimi K3 and Qwen 3.8 are raising fears that powerful AI is becoming easier and cheaper to copy, which could hurt companies like Anthropic. But commenters are fiercely split: some say Anthropic is trapped, while others insist people will still pay extra for the better overall experience.
Two flashy new artificial intelligence models, Kimi K3 and Qwen 3.8, have landed with a loud message: the fanciest AI may not stay exclusive for long. The article argues this is especially bad news for Anthropic, because if rival models perform almost as well and become publicly available, the premium-AI business starts looking shaky. But in the comments, readers didn’t just debate the business math—they went straight for the throat. One camp basically asked, “Why is Anthropic being singled out—how is OpenAI any different?” That turned the whole discussion into a cage match over who is actually vulnerable if cheaper, widely available AI gets good enough for most people.
Then came the Claude loyalists. One commenter flat-out said they’d still pay for Anthropic’s product because the experience feels better, comparing it to Apple vs Huawei/Xiaomi—a classic “yes, but the premium vibe matters” defense. Another waved off the panic entirely, saying power users happily pay high monthly fees if the tool saves them time, and that bargain-hunting is a loud but small crowd. On the opposite side, the doomers were having a feast: one argued Anthropic could get squeezed from both sides—beaten by cheap general-purpose tools for everyday tasks and by specialist AI for high-value jobs. And for extra spice, someone brought up the Figma board drama, reviving whispers about conflicts, timing, and whether this whole saga has more plot twists coming. In other words: the models dropped, but the comment section became the real battlefield.
Key Points
- •The article says Moonshot Labs’ Kimi K3 and Alibaba’s Qwen 3.8 were launched as state-of-the-art foundation models and are allegedly close to Anthropic’s Fable 5 in performance.
- •It states that both Kimi K3 and Qwen 3.8 are expected to have their model weights released publicly in the coming weeks.
- •The article argues that after training, the largest cost for foundation model companies is inference, driven mainly by compute and electricity.
- •It distinguishes between companies that lease infrastructure, companies that own data centers but buy electricity, and companies that own both data centers and power generation.
- •The article argues that firms with greater infrastructure ownership have more strategic flexibility because they can monetize infrastructure directly, including by hosting open models or leasing capacity.