MiniMax open-sources M3: 428B total params, 23B active, 1M-token context window

MiniMax M3 开源权重模型发布，已上架 HuggingFace

MiniMax uploaded M3 weights to HuggingFace, with the tech report and full weights expected in about 10 days. It's a 428B-total-param, 23B-active-param hybrid model using MiniMax sparse attention to push the context window to 1M tokens, plus native multimodal support. Coding and agent scores: SWE-Bench Pro 59.0%, Terminal Bench 2.1 66.0%, SWE-fficiency 34.8%, KernelBench Hard 28.8%, MCP Atlas 74.2%. MiniMax Code tool and API platform launched alongside. The post doesn't disclose training data, inference cost, or license terms — I'd hold off on usability judgments until the report drops.

Why it matters: MiniMax's first open-weight flagship release: 428B MoE with 23B active params and 1M context, with benchmark scores directly competing against DeepSeek and Qwen on agent/code tasks. Tech report still pending and weights just landed — clear info gaps — but the open-source move ...