Vercel Token Pass gives Pro teams $80 of Kimi K3 and GLM 5.3 credit for $30 a month

Vercel launched Token Pass on October 15, and it is in early access. Pro teams add $30 a month to their plan and get $80 of model credit for Kimi K3 and GLM 5.3. That works out to paying 37.5 cents per dollar of usage, which is worth a look if your team already codes mostly with these two open-weight models.

What you get

Each billing period you get $80 of credit, charged at the Token Pass provider's per-token prices listed on each AI Gateway model page. The featured models right now are Kimi K3 (moonshotai/kimi-k3) and GLM 5.3 (zai/glm-5.3), and Vercel plans to add more over time.

The credit belongs to the team, so everyone shares the same $80. Unused credit does not roll over. When it runs out, Token Pass requests fail until the next billing period starts, and there are no automatic overage charges. Your first charge is prorated for the rest of the current billing period, and the credit covers that same stretch.

Token Pass has zero data retention on by default. AI Gateway and the provider delete prompts and responses after each request, and your prompts are not used for training. Inference runs on provider infrastructure in the US and EU.

Who can subscribe

Pro teams can subscribe right away. Hobby teams need to upgrade to Pro first, teams on a Pro trial need to move to a paid Pro plan, and Enterprise teams cannot subscribe for now. Only team owners and members who can manage billing can subscribe or cancel.

Early access has a limited number of spots. When it is full, the Token Pass tab shows a waitlist instead, and joining the waitlist does not upgrade your plan or charge you.

How to use

1. Open the Token Pass tab in your team's AI Gateway dashboard and subscribe.
2. Create a Token Pass API key on the same tab. It only works with the featured models, so keep a regular AI Gateway key for anything else.
3. Run npx vercel@latest ai-gateway setup --key "your-token-pass-key" in a terminal. It finds the coding agents installed on your machine and configures them, including fx, Claude Code, Codex, Cline, and OpenCode.
4. Pick zai/glm-5.3 or moonshotai/kimi-k3 in your agent. Turn off any fast mode, because fast variants are not covered by Token Pass.

Token Pass renews monthly, and you can cancel from Billing Settings on the Token Pass tab. See the Vercel changelog post, the Token Pass docs, and the getting started guide.