← Library · Frontier
MiniMax M3 Released: Frontier Coding, 1M Context, and Native Multimodality in an Open-Weight Model
MiniMax has officially released M3, an open-weight model that integrates frontier-level coding and agentic capabilities, a 1M token context window, and native multimodality. It achieves this by using MiniMax Sparse Attention (MSA), a new attention architecture, which contributes to significant speedups and efficiency gains. Testing shows M3 surpasses GPT-5.5 and Gemini 3.1 Pro on coding benchmarks like SWE-Bench Pro, and outperforms Opus 4.7 on SVG generation.
Why it matters
M3 is the first and only open-weight model to combine these three frontier capabilities, offering developers a powerful and flexible tool for complex AI applications with impressive performance benchmarks.
Learn one new AI thing every day.
Daily Deck sends you seven plain-English cards like this every morning. Free.
Start free