MiniMax M3 open-source model matches GPT-5, Claude Opus on benchmarks

MiniMax has released its M3 model, an open-source model that rivals top closed-source competitors in long context, multimodal, and coding capabilities. Early tests show M3 successfully replicating research papers, generating interactive web pages based on complex prompts, and even creating a webpage detailing NVIDIA CEO Jensen Huang's AI ecosystem partners. The model's efficiency is highlighted by its significantly reduced computational cost per token and accelerated decoding speeds, with MiniMax also launching a complementary coding tool, MiniMax Code, designed to work alongside M3. AI

IMPACT Sets a new benchmark for open-source models, potentially driving down costs and increasing accessibility for advanced AI capabilities.

RANK_REASON This is a new model release from a frontier lab (MiniMax) with detailed performance metrics and comparisons to other leading models. [lever_c_demoted from frontier_release: ic=1 ai=1.0]

Read on 量子位 (QbitAI) →

• Claude
• Gemini
• GitHub
• GPT-5
• Guillermo Rauch
• Jensen Huang
• MiniMax
• MiniMax Code
• Nous Research
• NVIDIA
• Opus
• Vercel

• model release
• product

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →