Verified, then logged.
Sun 22 Nov 2026 · 00:34 KST / 15:34 UTC한국어 | ENSearch /

02 / IT · Analysis

Vercel expands AI Gateway generation and decision models

Document generation and decision outputs serve different tasks, while decision calls count towards the same budgets as other model requests.

An ethernet cable is plugged back into a laptop on a developer's desk at night while a small workstation keeps running behind it with amber activity lights. AI-generated image

Vercel announced broader support for generation and decision models through AI Gateway, without naming a licence in its notices.

The announcements distinguish models that create documents and other deliverables from a model that returns choices, scores and probabilities. That distinction also extends to cost management: decision calls count towards budgets alongside other model requests.

Vercel's notices specify the model names developers use to make requests. They do not disclose weight distribution or licence names, so the announcements describe access through AI Gateway rather than terms for downloading the models.

Tasks for generation models

Vercel presents Gemini 4 Hydrogen as the successor to Gemini 3.8 Flash, with latency remaining in the same class.

According to Vercel, the model improves coding and tool use and handles longer engineering tasks. It produces documents, spreadsheets and presentations in fewer steps.

Vercel describes Inkling 2 Small as accepting images and returning text. It also presents screenshot-based checks of generated code as a way to guide revisions.

Outputs from the decision model

Vercel presents Gemini Decision 1 as a model that returns choices, scores and probabilities for true-or-false outcomes. It describes this process as operating without generating output tokens, the units of generated text.

According to Vercel, supported inputs depend on the interface used to call the model. AI SDK 7 supports PDF, audio and video inputs; the TypeSafe-compatible API and Decisions API support text and images.

Cost management through one gateway

Vercel states that AI Gateway applies provider pricing without a markup and charges no platform fee for inference, the process of running a model. Those conditions also apply to requests made with a user's own API key.

According to Vercel, decision calls appear in logs and count towards budgets alongside other model requests. Cost tracking therefore brings the requests together, while developers still need to choose an interface that supports their intended inputs.