MiniMax M3 Now Live on Vercel AI Gateway
Vercel has added MiniMax M3 to its AI Gateway, and developers can start using it right now. The integration requires no separate MiniMax account and comes with zero pricing markup on inference.
M3 is MiniMax’s most capable model to date. It features a 1-million-token context window, one of the largest available, and native multimodality built on MiniMax Sparse Attention (MSA). In practical terms, that means the model can process both text and images within a single prompt. It also handles multi-turn collaboration well, making it strong for agentic tasks.
Specifically, M3 shows improvements in software engineering, terminal-based tool use, and agentic web browsing. These capabilities make it well-suited for developer workflows that go beyond simple prompting.
To use MiniMax M3 Vercel AI Gateway integration, developers simply set the model string to minimax/minimax-m3 inside the Vercel AI SDK. For multimodal input, images are passed alongside a text prompt in a standard message format.
Vercel’s AI Gateway provides a single, unified API endpoint for calling multiple AI models. It handles usage tracking, cost monitoring, retries, failover, and performance optimizations, all without charging a platform fee on inference. The Gateway also supports Zero Data Retention, custom reporting, and dynamic provider sorting by latency and cost. Bring Your Own Key (BYOK) requests are also fully supported.
Developers who want to explore MiniMax M3 can try it directly in the Vercel model playground or review the full AI Gateway model leaderboard to compare options. With MiniMax M3 now live on Vercel AI Gateway, accessing frontier-level context and multimodality has become considerably simpler for teams already building on Vercel.
