MiniMax M2.7 Available on AI Gateway

MiniMax's M2.7 model is now accessible through Vercel's AI Gateway, offering developers a significant upgrade for software engineering and agentic tasks. The model is available in two variants:

• Standard: Full performance for standard use cases
• High-speed: 2x throughput at ~100 tokens per second, optimized for latency-sensitive applications

Key Capabilities

M2.7 introduces native support for advanced workflows:

• Multi-agent collaboration and complex skill orchestration
• Dynamic tool search for intelligent agentic workflows
• Enhanced production debugging and end-to-end project delivery
• Improvements in professional office tasks and software engineering

Getting Started

To use M2.7 with the Vercel AI SDK, set the model parameter to either minimax/minimax-m2.7 or minimax/minimax-m2.7-highspeed. The example below demonstrates using the high-speed variant for analyzing production alerts and submitting fixes:

import { streamText } from 'ai';

const result = streamText({
  model: 'minimax/minimax-m2.7-highspeed',
  prompt: `Analyze the production alert logs from the last hour,
     correlate them with recent deployments, identify the
     root cause, and submit a fix with a non-blocking
     migration to restore service.`,
});

AI Gateway Features

M2.7 joins other models on Vercel's unified AI Gateway, which provides:

• Unified API for calling multiple models
• Usage tracking and cost monitoring
• Configurable retries and failover strategies
• Built-in observability and provider routing
• Bring Your Own Key (BYOK) support for cost optimization