Vercel offers Glyph Cluster reasoning model free in stealth on AI Gateway

Glyph Cluster arrives in stealth

Vercel has released Glyph Cluster, a reasoning model built for coding and long-context analysis, as a stealth model on its AI Gateway platform. According to the Vercel blog, the model is free to use for the duration of the stealth period, though access is limited to teams on Pro and Enterprise plans that have purchased AI Gateway credits.

Vercel positions the model around multi-step work rather than single-shot answers. The company says it handles planning, synthesis, quantitative reasoning, and comparing material across large inputs. For development workflows specifically, the announcement lists reviewing code, explaining failures, debugging issues, proposing changes, and reasoning through competing implementation options.

Features and limitations

Glyph Cluster supports function calling, so agents can wire the model's reasoning into tools and systems the developer defines. Responses stream as they are generated, which is the behaviour coding agents such as Claude Code, Codex and Cursor already expect from a model endpoint.

The limitations are worth reading closely. The model takes text input only, with no image or file input. Tool use is restricted to function tools you define yourself, and structured outputs are not supported. Zero Data Retention is not available for this model, and Vercel states plainly that prompts and responses sent through it may be used for training and model improvement. For teams with strict data policies, that means anything routed through Glyph Cluster during this period should be treated as potentially contributing to the model's development.

How to call it

According to the announcement, there are three routes in: the AI SDK, the OpenAI-compatible Chat Completions and Responses APIs, or a coding agent connected to AI Gateway. The model name to use is stealth/glyph-cluster.

For Claude Code, Codex, Cursor and similar tools, Vercel says to install the latest CLI with npm i -g vercel@latest, run vercel ai-gateway setup, and then select the stealth model inside the agent. The company also points to its model playground for trying the model directly.

The gateway around it

AI Gateway is Vercel's unified API layer for calling models across providers, with usage and cost tracking, retries, failover, and performance optimizations that Vercel claims deliver higher-than-provider uptime. It also offers built-in custom reporting, Zero Data Retention support where a model allows it, budgets for API keys, and routing rules.

Vercel notes that the gateway passes through provider pricing with no markup and charges no platform fee on inference, including on Bring Your Own Key requests.

Why it matters

A free stealth window gives paying teams a chance to benchmark a new reasoning model on real workloads at no inference cost, limited only by the plan and credit requirements they already meet. The eligibility gate makes this a perk for existing Vercel customers rather than an open beta, which keeps evaluation traffic contained while the model is still being tuned.

The trade-off is data. With no Zero Data Retention and explicit language about training use, the free period effectively exchanges cost for data, so teams would be wise to route lower-sensitivity tasks through it first. On the convenience side, little stands in the way of a quick trial: function calling and streaming are supported out of the box, the endpoints are OpenAI-compatible, and setup in popular coding agents is a two-command CLI step. If Glyph Cluster proves competitive on code review and long-context analysis, Vercel has lowered nearly every barrier to trying it except the data one.