MiniMax H3 Image and Highspeed models land on Vercel AI Gateway with 2K and 4K output

MiniMax H3 Image models arrive on Vercel AI Gateway

Vercel has added MiniMax H3 Image and MiniMax H3 Image Highspeed to AI Gateway, the company's unified routing layer for AI models. According to the Vercel blog, developers can now generate images from text prompts and edit existing pictures using a single API key, with no need to set up a separate MiniMax account.

Generation and editing in one workflow

The two models cover a full create-and-refine loop. A developer can start with a text prompt to produce a product shot or an illustration, then send the resulting image back to the model together with new instructions, for example to alter colors, replace a subject, or rework the composition.

Resolution is a central part of the announcement. Vercel says both models accept custom dimensions and aspect ratios, with output sizes reaching 2K and 4K. The Highspeed variant offers quicker generation while keeping the same generation and editing options, which makes it the pick when turnaround time matters most.

Integration details

The models are addressed by the identifiers minimax/minimax-h3-image and minimax/minimax-h3-image-highspeed. Both are also reachable through the OpenAI-compatible routes /v1/images/generations and /v1/images/edits, so code already written against OpenAI's image endpoints can be pointed at the MiniMax models with minimal changes. Vercel also lists the models in its model playground for anyone who wants to experiment before writing any code.

Billing and the gateway itself

AI Gateway is positioned as a single entry point for calling models across providers, monitoring usage and spend, and managing access. Requests and costs surface in the Vercel dashboard. On pricing, Vercel states that inference is billed at the provider's own rates, with no markup or platform fee added on top, and points readers to its AI Gateway pricing page for the specifics.

Why it matters

Image generation has splintered across a growing number of providers, each with its own account setup, request format and billing relationship. Pulling MiniMax's image models into AI Gateway collapses that overhead into one key, one request surface and one usage view.

Two details stand out for practical adoption. The OpenAI-compatible endpoints keep switching costs low: teams already using the standard image API shape can evaluate MiniMax H3 Image without a rewrite. And 2K and 4K output moves generated imagery closer to production-grade assets for use cases like e-commerce listings and marketing material, where low-resolution renders often fall short. With pricing passed straight through from the provider, routing via Vercel adds convenience without inflating the per-image bill, a combination that lowers the barrier to testing these models against the other image models already available on the gateway.