Vercel offers MiniMax M3.1 Flash free on AI Gateway until October 22
The promotion applies only when MiniMax serves the request. Vercel says requests pinned to MiniMax will move to standard token pricing after the offer ends.
By Benchdict AI DeskPublished Oct 20, 20264 min read
Vercel announced on October 19 that MiniMax M3.1 Flash is available free through AI Gateway until October 22, but only for requests served by MiniMax. The company says requests that remain pinned to MiniMax will become billable at its standard rate after the promotion.
Key facts
- Free-offer deadline
- Until October 22, according to Vercel; exact cutoff unspecified
- Serving-provider condition
- Only requests served by MiniMax receive the free rate
- Standard token prices
- MiniMax lists $0.15 input, $0.60 output and $0.015 cached input per million tokens
- Gateway model identifier
- minimax/minimax-m3.1-flash
Free access depends on the serving provider
AI Gateway is Vercel's unified API for calling models, and it passes through each provider's pricing. Vercel instructs developers to prioritize MiniMax using order in AI Gateway's provider options to obtain the promotional rate; requests set up this way are the ones the offer calls pinned to MiniMax. That makes the serving provider—not simply the model identifier—the boundary of the offer. Selecting M3.1 Flash alone does not guarantee that a request qualifies for free usage.
For developers routing prototype workloads through a gateway, provider selection is also a billing consideration. Vercel's announcement does not establish what happens to charges when a request falls back to another provider, or provide a complete configuration for preventing that outcome.
According to Vercel, the model can be called through the AI SDK, Chat Completions, Responses and Anthropic Messages APIs, as well as connected coding agents. Its coding-agent setup instructions specify the latest Vercel CLI and the ai-gateway setup command.
The promotion ends; pinned requests become paid
Vercel says requests pinned to MiniMax will be billed at MiniMax's standard rate after the promotion. MiniMax publishes the following flat token prices:
| Token category | MiniMax's listed price per million tokens |
|---|---|
| Input | $0.15 |
| Output | $0.60 |
| Cached input | $0.015 |
Vercel says AI Gateway passes through provider pricing without markup or an inference platform fee, including for bring-your-own-key requests. Separately, its API-key documentation says keys can have optional spending budgets. Those budgets are spending controls, not confirmation of the promotion's rate limits or capacity.
For teams leaving experimental integrations running, the announced transition is therefore from promotional usage to token billing—not an announced shutdown of access.
From a coding-tool preview to broader access
DataNorth previously reported the preview's release inside MiniMax Code in September. MiniMax's October 19 announcement describes availability through its API, M Plan and MiniMax Code, while Vercel adds a gateway route for developers using its supported interfaces.
MiniMax says overall agentic-task usage costs are 70% lower than the predecessor's. That is a company-reported comparison without an absolute cost baseline or independent replication here; it does not show a cost advantage over rival models. The published token schedule is a firmer basis for understanding post-promotion billing than that task-level claim.
There is also a material distinction between hosted access and downloadable weights. Vercel describes M3.1 Flash as open-weight, but MiniMax says the weights will be released later. For developers considering self-hosting rather than a hosted API, the latter statement leaves availability unresolved.
Not yet confirmed
- Vercel does not specify the exact cutoff time, timezone or whether the promotion includes all of October 22.
- Promotion quotas, rate limits, concurrency limits, regional restrictions and account or payment-method eligibility remain unconfirmed.
- Charges for fallback requests served by another provider, and a complete configuration to prevent such charges, remain unconfirmed.
- MiniMax's reported task-cost reduction lacks independent confirmation and full measurement conditions.
- The weight-release date, download location, license and self-hosting requirements remain unconfirmed.
- Model-specific retention guarantees and treatment of sensitive code remain unconfirmed.
What to watch
- Vercel's clarification of the October 22 cutoff and transition to standard billing.
- Publication of promotion capacity limits and fallback-billing details.
- MiniMax's release of model weights and licensing terms.
Sources
- MiniMax M3.1 Flash now available on AI Gateway for freeVercel Blog, Oct 19, 2026.
- MiniMax M3.1 FlashHacker News (front page), Oct 19, 2026.
- AI Gateway API Keysvercel.com, Oct 20, 2026.
- MiniMax released M3.1-Flash-Preview - DataNorth AIDataNorth AI, Sep 28, 2026.
Benchdict AI Desk, AI-assisted reporting desk. Stories follow our editorial standards; corrections are logged above and on the corrections page.