GLM 5.2 free for eve agents through August 27 via Blackbox on AI Gateway
Z.ai's GLM 5.2 open-weights coding model is now available at no cost to developers building with eve agents, served through Blackbox AI on Vercel's AI…
By Dillip Chowdary • Aug 16, 2026 • Source: Vercel Blog
What happened
Z.ai's GLM 5.2 open-weights coding model is now available at no cost to developers building with eve agents, served through Blackbox AI on Vercel's AI Gateway. The offer runs through August 27, giving builders a limited window to run a model with a 1M-token context window without incurring inference charges.
This article covers what shipped, how the integration works, and what developers need to do — whether starting a new agent or updating an existing one. It is aimed at engineers already working in the eve ecosystem or considering it for their next coding assistant or autonomous agent project.
What shipped
How it works
Z.ai's GLM 5.2 is an open-weights model designed for coding tasks, carrying a 1M-token context window. Vercel has partnered with Blackbox AI to serve it through AI Gateway, and the arrangement makes GLM 5.2 the default model for all newly created eve agents until August 27. The zero-cost period covers the full context window, meaning builders are not paying for prompt or completion tokens during this window. Blackbox AI acts as the serving layer between AI Gateway and the underlying model weights, so the integration is handled at the infrastructure level rather than requiring developers to configure a separate API key or provider.

What changed for builders
Why it matters
Advertisement
Tech Pulse Daily
Get tomorrow's pulse first
Join engineers who read Tech Pulse before stand-up. Free, weekday mornings.
The most immediate change is the default model selection for new eve agents. Any agent initialized during this period will use GLM 5.2 out of the box rather than requiring a manual model selection step. For existing agents, the benefit is available but not automatic, so developers need to take an explicit action to opt in. The 1M-token context window is the headline capability shift: coding tasks that previously required chunking large codebases or truncating context can now pass significantly more source material in a single call, which affects how retrieval, summarization, and multi-file editing tasks are structured.
How to install or upgrade
To start a new agent using GLM 5.2 as the default, run npx eve@latest init my-agent from a terminal. The init command pulls the latest version of the eve CLI and scaffolds a new agent project with GLM 5.2 already configured as the model. Existing agents need to update to take advantage of the free period. The exact update path depends on how the agent was originally configured, but upgrading the eve package to its latest version and verifying the model field in the agent configuration is the expected first step. Because the offer is tied to the AI Gateway routing layer, no separate Blackbox AI account setup should be required for agents running through Vercel's standard AI Gateway path.
Who is affected
Gotchas and compatibility
The free access period ends on August 27, so any agent relying on GLM 5.2 as its default will need a model migration plan or cost budget in place before that date. What happens after August 27 — whether the model becomes paid, is removed from AI Gateway, or continues under different terms — is not specified in the announcement, so builders should treat this as a time-bounded trial rather than a permanent configuration. Existing agents are not automatically switched to GLM 5.2, meaning teams that want to test the 1M-token context window on live workloads need to explicitly update their configuration rather than expecting a passive rollout from Vercel or Blackbox AI.
What to watch next
What to watch next
The key variable to track is what happens to GLM 5.2's availability and pricing on AI Gateway after August 27. If Blackbox AI continues offering the model beyond that date under a paid tier, builders who have tuned workflows around the 1M-token context window will want to compare that cost against alternative long-context models available through AI Gateway. Z.ai's open-weights release means the model can also be self-hosted, which may become relevant for teams with latency or data-residency requirements once the free period ends. Watch for updates from Vercel on whether additional open-weights models enter similar promotional windows through the Blackbox AI partnership.
Advertisement