MiniMax M3, M2.7 Go Free on AI Gateway

This is a useful temporary window for developers evaluating MiniMax in real agent workloads, but the promotion is best treated as a trial—not production pricing.

• –M3’s million-token context and multimodal input suit repository-scale coding, screenshots, and long-running agent tasks.
• –M2.7 is positioned for software engineering, debugging, tool use, and autonomous workflows.
• –AI Gateway provides a unified API, provider routing, failover, and observability across deployments.
• –Standard pricing resumes after the 14-day campaign, so benchmark latency, rate limits, and token usage before committing.
• –Free inference is an aggressive acquisition tactic that makes model switching and side-by-side evaluations nearly frictionless.