News
Chinese Open Model GLM-5.2 Sees Fastest Adoption on Vercel This Year
Chinese lab Z.ai's GLM-5.2 model posted an 80-fold increase in customers and a 27-fold jump in query volume in its first full week on Vercel AI Gateway, delivering results close to Western flagship models at a significantly lower price.
GLM-5.2, a model from Chinese lab Z.ai, posted an 80-fold increase in customers and a 27-fold increase in daily token processing volume in its first full week on Vercel AI Gateway. According to the platform itself, this is the fastest adoption of a single model Vercel has recorded in 2026.
GLM-5.2 is the successor to the earlier GLM-5.1 and part of a broader push by Chinese AI labs into the open-model coding market. The model is available both through the Z.ai API and for download in GGUF format on Hugging Face, allowing it to be run locally without subscription fees.
Growth on Vercel
Vercel AI Gateway is a middleware layer that lets developers using tools such as Claude Code, Cursor, or the Vercel AI SDK plug in different language models without changing their application code. The sharp rise in the number of customers using GLM-5.2 on the platform suggests software companies are actively testing it as a cheaper alternative to OpenAI and Anthropic models in day-to-day coding work.
The significance of this single integration becomes clearer against a broader trend: models from Chinese labs now account for 30 to 46 percent of token usage flowing through US enterprise developer platforms. A year ago, such a share would have been hard to imagine in a segment dominated by OpenAI, Anthropic, and Google.
Results Close to Western Models
On the SWE-bench Pro benchmark, which measures a model's ability to independently solve real-world coding tasks, GLM-5.2 scored 62.1 percent. That's comparable to leading Western closed models and a clear jump from the previous version, GLM-5.1, which scored 63.5 points on Terminal-Bench 2.1 versus 81.0 points for the new version. On app-building tasks, the number of solved cases rose from 21 to 48 out of 70 possible.
The model also earned praise from Vercel itself. Guillermo Rauch, the platform's CEO, publicly described his impressions after testing GLM-5.2's coding capabilities.
I was genuinely impressed, almost shocked - Guillermo Rauch, CEO, Vercel
The Math for Businesses
For software companies, what matters isn't just benchmark scores but the ratio of quality to price. At $1.40 per million input tokens and $4.40 per million output tokens, GLM-5.2 costs noticeably less than flagship closed models, and the ability to run it locally without a subscription further lowers the total cost of deployment at larger organizations. This combination of near-premium performance, lower pricing, and open distribution is starting to factor into the calculations of IT departments that must justify AI subscription spending to executives scrutinizing every dollar of infrastructure cost.
What's Next
For Polish software companies and dev shops using tools like Cursor or coding agents built on Vercel AI Gateway, the arrival of a cheaper, fully competitive alternative creates a real opportunity to negotiate costs or shift some tasks to a cheaper model without losing quality. The rapid pace of GLM-5.2's adoption also shows that the barrier to switching between language model providers is getting lower as the performance gap between open and closed models narrows. Sources: Cryptobriefing (cryptobriefing.com), Tech Reader (tech-reader.blog)
