China's GLM-5.2 Hits 1/6 the Cost, 1 Point Off Opus 4.8 — Anthropic's Paywall Never Stood a Chance 🐉

Z.ai released GLM-5.2, an open-source system with 750 billion parameters and a 1-million-token context window, running entirely on domestic Chinese chips and priced at roughly one-sixth the cost of comparable US frontier models. The model now sits within a single percentage point of Anthropic's Opus 4.8 on a closely watched agentic evaluation, and on an internal app-development suite it scored 48 out of 70, up from GLM-5.1's 21, compared with 56 for Claude Fable 5.

The launch landed a day after Anthropic disabled global access to its most advanced models, including Fable 5 and Mythos, and the same week OpenAI moved to limit access to GPT-5.6 following a separate government request. Z.ai co-founder Tang Jie called the Anthropic suspension "deeply regrettable" and said frontier intelligence should not belong to a few people or be subject to sudden rule changes. Z.ai, formerly known as Zhipu AI, framed its open-weights approach as a more stable institutional option for enterprise buyers.

Pricing is where the comparison sharpens. DeepSeek V4 Pro charges $3.48 per million output tokens against the $50 that Anthropic's Fable 5 charged for the same volume, and GLM-5.2 enters the same low-cost tier. Adoption metrics tracked on OpenRouter reflect the shift among enterprise customers reassessing vendor relationships in light of US access restrictions.

Markets reacted within hours. Z.ai shares jumped more than 30% in Hong Kong trading, extending gains to over 800% since its January debut. JP Morgan projects Z.ai revenue to expand by more than 534% this year, with profitability arriving by 2028. Knowledge Atlas Tech Joint (Zhipu AI) led the move on the Hong Kong board following the GLM-5.2 release on June 19, 2026.