MiniMax M3 is now accessible via Vercel AI Gateway using the model identifier minimax/minimax-m3 in the AI SDK. It is MiniMax's first model with a 1M-token context window and native multimodality, built on MiniMax Sparse Attention architecture.
M3 targets software engineering workflows, terminal-based tool use, and agentic web browsing, with tuning specifically for multi-turn collaboration. The multimodal input accepts images alongside prompts directly through the API.
Vercel AI Gateway routes the request through a unified API that handles usage tracking, retries, failover, and dynamic provider sorting by latency and cost, all at provider pricing with zero markup and no platform fee, including on Bring Your Own Key requests. The model leaderboard and playground give you a concrete baseline before committing.
[READ ORIGINAL →]