MiniMax M3 is an open-weight multimodal foundation model with a 1 million token context window. It edges out GPT-5.5 on vendor-reported coding benchmarks while costing significantly less. The model uses MSA (MiniMax Sparse Attention) architecture for efficient long-context processing.
AI Agents
1 min read
Jun 01, 2026
MiniMax M3: The $0.30 AI Model That Edges Out GPT-5.5
Chinese AI company MiniMax released M3 with a 1 million token context window, frontier coding performance, and pricing 16x cheaper than GPT-5.5.
A
Abdallah Mohamed
Senior Full-Stack Engineer
