DeepSeek V4.1 Flash joins Vercel's Token Pass alongside Kimi K3 and GLM 5.3

DeepSeek V4.1 Flash joins Token Pass

Vercel has added DeepSeek V4.1 Flash to Token Pass, the monthly model subscription offered to Pro teams. According to a changelog post on the Vercel blog dated October 20, the new model now sits alongside Kimi K3 and GLM 5.3, which were already part of the lineup, and Vercel says further models are on the way. Each model in the plan is served through its preferred provider, and the DeepSeek option runs on the newest weights, version 1017.

What Token Pass includes

Token Pass costs $30 per month on Pro teams and comes with $80 of monthly credit, with all usage routed through Vercel's AI Gateway. Zero data retention is switched on by default, so prompts and completions are not kept as requests pass through the gateway.

The subscription is pitched at coding-agent workflows. Vercel names fx, Pi, Cline and Command Code among the agents it works with, and developers can also build their own integrations using the AI SDK. In practice, that means a single subscription can back several different tools without managing separate accounts for each model.

How to enable it

Subscribers can add DeepSeek V4.1 Flash from the Token Pass page inside their team's AI Gateway dashboard. Vercel points anyone who needs configuration details to its Token Pass documentation.

Why it matters

The addition gives Token Pass subscribers a third model family to spend their credit on. Kimi K3, GLM 5.3 and DeepSeek V4.1 Flash occupy the same broad category — large models aimed at coding and agentic work — and being able to swap between them under one credit pool removes much of the friction of comparing them on real tasks. A developer who finds one model underperforms on a particular codebase can point their agent at another without opening a new account or setting up separate billing.

The economics are also unusual. A fixed $30 outlay buys $80 of credit, a gap that presumably reflects Vercel's ability to source inference below retail rates. For subscribers, the effect is to turn per-token pricing into a predictable monthly line item, and every model Vercel adds to the pool increases what that $30 buys without changing the price.

There is a distribution angle as well. DeepSeek, Kimi and GLM reaching developers through a bundle assembled by a deployment platform, each routed via its preferred inference provider, illustrates how model access is consolidating: fewer direct signups with individual labs, more flat-fee subscriptions layered on top of gateway routing. For Vercel, each addition also keeps AI Gateway and the surrounding AI SDK ecosystem sticky among the agent-tooling audience it already serves.