Vercel Makes MiniMax M3 and M2.7 Free Through AI Gateway.

Vercel is temporarily removing the usage cost for two MiniMax models on AI Gateway, giving developers a limited window to test M3 and M2.7 without model charges. The free routes are available through Sunday, September 6, via infrastructure provider GMI Cloud.

The company announced the offer in its changelog on August 25. While the promotion lowers the barrier to evaluating the models, developers need to pay attention to the model IDs: applications using the dedicated free identifiers will start returning errors when the offer ends.

Developers can access the models through Vercel AI Gateway using one of two identifiers:

• minimax/minimax-m3-free
• minimax/minimax-m2.7-free

Both routes send requests to GMI Cloud. In an application using Vercel's AI SDK, streaming a response from M3 requires only the free model ID:

import { streamText } from 'ai';

const result = streamText({
  model: 'minimax/minimax-m3-free',
  prompt: 'Summarize the incident report.',
});

That makes the promotion useful for short evaluation cycles, prototypes and internal experiments. A small business could, for example, test document summarisation or an AI assistant before deciding whether the model belongs in a production workflow.

The key limitation is continuity. Vercel says the -free model IDs will return an error after the promotional period, rather than automatically switching to a paid route. Any prototype expected to remain online beyond September 6 will therefore need a code or configuration change.

For continued access, Vercel recommends replacing the promotional identifier with the standard model ID and explicitly placing GMI Cloud first in the gateway's provider order. For MiniMax M3, the configuration looks like this:

const result = streamText({
  model: 'minimax/minimax-m3',
  prompt: 'Summarize the incident report.',
  providerOptions: {
    gateway: {
      order: ['gmicloud'],
    },
  },
});

This approach prefers GMI Cloud while avoiding a dependency on an identifier designed to expire. Teams adopting the free route should treat the transition as part of deployment planning, especially if the model is connected to a customer-facing feature or an automated workflow.

Vercel did not provide standard post-promotion pricing in the announcement. Developers should review the applicable AI Gateway model and provider terms before moving a test into production.

The update fits Vercel's broader effort to assemble infrastructure for AI applications around its deployment platform. That includes model routing through AI Gateway and execution tooling such as the Vercel Run SDK, which isolates untrusted code from AI agents. For smaller development teams, a gateway can reduce the integration work involved in testing different models and providers, but it does not eliminate the need to monitor costs, availability and model-specific behaviour.

The practical takeaway is straightforward: the free identifiers are suitable for time-boxed testing, not permanent configuration. Developers who decide to keep MiniMax M3 or M2.7 should switch to the standard IDs before the deadline and verify the resulting provider and billing setup.

What is Vercel MiniMax M3 and M2.7?

MiniMax M3 and M2.7 are AI models that developers can access through Vercel AI Gateway. Vercel is offering dedicated free routes for both models through GMI Cloud for a limited promotional period.

How long are MiniMax M3 and M2.7 free on Vercel AI Gateway?

The free access runs through Sunday, September 6. Vercel says the model IDs ending in -free will return an error after that period ends.

How can developers keep using the models afterward?

Developers must replace the promotional ID with the corresponding standard model ID, without the -free suffix. Vercel also recommends placing gmicloud first in the AI Gateway provider order if the application should continue to prefer GMI Cloud.