Zhipu AI Drops a 744B Open-Weight Model That Beats GPT-4.5 on Coding — as DeepSeek Hits $4.4B Valuation
China's open-weight offensive is no longer a sideshow. Zhipu AI has released a 744-billion-parameter model that reportedly outperforms GPT-4.5 on coding benchmarks, while DeepSeek has become the country's most valuable AI startup after a $4.4 billion raise. The implications for U.S. frontier labs' pricing strategies are serious.
China's AI ecosystem produced two headline-worthy developments this week that, taken together, represent the clearest signal yet that the U.S. frontier labs' moat is narrower than commonly assumed. As @yo123dawg reported, Zhipu AI has released a 744-billion-parameter open-weight model that beats OpenAI's GPT-4.5 on coding benchmarks, while DeepSeek has closed a funding round that values the company at $4.4 billion — making it China's most valuable AI startup.
The coding benchmark results, if they hold up under independent evaluation, are striking for their specificity. GPT-4.5 has been OpenAI's flagship commercial model for enterprise coding workflows, and a Chinese open-weight model matching or exceeding it on those exact tasks undermines the core value proposition that has justified premium API pricing. Open-weight means any developer or company can download, fine-tune, and deploy the model without per-token fees — a structural cost advantage that no amount of product polish can fully counter.
Get our free daily newsletter
Get this article free — plus the lead story every day — delivered to your inbox.
Want every article and the full archive? Upgrade anytime.
No spam. Unsubscribe anytime.