Vercel adds Z.ai's GLM 5.3 Flash to Token Pass with 1M-token context

Vercel has expanded the model selection in Token Pass, its subscription plan for AI usage, by adding GLM 5.3 Flash. According to the company's changelog, the newcomer joins Kimi K3, GLM 5.3 and DeepSeek V4.1 Flash in the lineup, and Vercel indicates that further additions are planned.

A multimodal coding model with a million-token window

GLM 5.3 Flash is built by Z.ai and, per Vercel, is a native multimodal coding model, meaning it can work with more than plain text as input. Its standout specification is a context window of one million tokens, which in principle is large enough to hold a whole codebase, extensive documentation or a long agent session inside a single request.

That matters for coding agents in particular. The tools Token Pass targets repeatedly re-read project files while planning and applying changes, and context size determines how much of a repository the model can consider at once. A bigger window reduces the need for chunking, summarising or retrieval workarounds just to keep the model oriented.

What the subscription includes

Token Pass is priced at $30 per month on Pro teams and bundles $80 of monthly credit, so subscribers effectively receive more usage credit than they pay in subscription fees. All traffic runs through Vercel's AI Gateway, the company's centralized service for routing requests to model providers.

According to Vercel, the subscription can be used with any coding agent, and the changelog names fx, Pi, Cline and Command Code among the compatible options. Teams that prefer their own tooling can build against the AI SDK, Vercel's framework for constructing AI applications.

Vercel also notes that Zero Data Retention is switched on by default, so prompts and outputs handled through the subscription are not kept by the provider, a relevant detail for teams working with proprietary code.

How to enable it

Turning the model on is a configuration step rather than a code change. Vercel directs users to the Token Pass page within their team's AI Gateway dashboard to add GLM 5.3 Flash, with the Token Pass documentation covering the rest of the setup for agents and custom integrations.

Why it matters

The addition says something about how model access is being packaged. Rather than maintaining separate accounts and billing relationships with individual providers, Token Pass subscribers get a growing roster, now four models, under one subscription, routed through a single gateway with one monthly credit pool.

The vendor mix is also notable: models from Z.ai, Kimi and DeepSeek sit side by side in the same plan, suggesting Vercel is curating on capability and price rather than tying itself to one supplier. The economics, $80 of usage for a $30 fee, are aggressive enough to put pressure on conventional per-token billing for developers already working inside the Vercel ecosystem.

Finally, the framing is agent-first. Token Pass is pitched at autonomous coding tools and custom AI SDK builds rather than chat interfaces, a bet that serious model consumption now happens inside agents working over entire repositories, which is exactly the workload a one-million-token context window is built to serve.