Zhipu Technology (智谱) has enacted the most significant price increase by a domestic large-language-model vendor since the start of 2026, raising subscription and API fees for its GLM Coding Plan as it simultaneously launched the new GLM-5 model overseas. The company announced on February 12 that package prices would rise broadly, existing subscribers would keep their current rates, and the first-purchase discount would be removed while seasonal and annual discounts remain. The move takes effect immediately and, according to documents released by Zhipu, responds to sharply accelerating developer demand and sustained high-load usage of its code-generation and programming-assistant products.
The scale of the increase is material. The company said domestic package prices would climb by at least 30%, while a version of its overseas roll-out showed even larger adjustments: Coding Plan subscriptions up 30–60% and API call prices rising between roughly 67% and 100%. Zhipu justifies the change as necessary to underwrite heavier investment in compute capacity, model optimisation and operational resilience as usage outstrips the company’s original infrastructure plans. To blunt customer pain, the firm has frozen prices for existing subscribers and said further adjustments will be guided by market feedback.
Markets responded quickly: Hong Kong AI concept shares linked to Zhipu jumped on the news, reflecting investor appetite for demonstrable monetisation. The company’s simultaneous overseas release of GLM-5 — rather than a domestic-only debut — amplified attention, framing the price move as part of a global commercial strategy rather than a local test. The rollout has not been entirely smooth; online critics allege technical and provenance questions about GLM-5, underscoring reputational risks that can accompany rapid product expansion.
This episode matters because it signals a structural shift in China’s big-model sector. After years of low-cost, subsidy-driven user acquisition, at least some vendors are now charging rates that reflect rising unit costs and product value. Industry data cited by Zhipu and market trackers align: semiconductor markets remain robust and AI compute demand is a leading growth driver, squeezing suppliers and lifting hardware and cloud costs. Vendors that once prioritized market share through aggressive pricing increasingly confront trade-offs between growth, service quality and sustainable margins.
The near-term consequences are straightforward. Higher API and subscription prices will raise the threshold for small-scale experimentation and shift some usage toward enterprise customers with predictable budgets or to vendors offering cheaper, lower-capacity tiers. That could benefit cloud and chip suppliers as enterprise contracts expand, while producing churn risk among cost-sensitive developers and start-ups. Zhipu’s decision to grandfather existing users is designed to mitigate reneging, but broader sectoral price normalisation would test developers’ willingness to pay and the elasticity of token consumption.
Strategically, Zhipu’s move is a bellwether. It demonstrates growing commercial confidence among China’s AI start-ups and highlights the centrality of compute investments to product reliability. It also raises governance and competitive questions: rivals will weigh whether to follow, regulators and customers will scrutinise service-level assurances, and global expansion will expose Chinese models to reputational and IP-related scrutiny. For foreign buyers and partners, the episode is a reminder that commercialisation — not just capability parity — will determine which models win long-term adoption.