MiniMax M3 Released: 1M-Context Open-Weight Coding Model Beats GPT-5.5 on SWE-Bench Pro

MiniMax·high signal

MiniMax launched M3 on June 1, 2026 with a new MSA (MiniMax Sparse Attention) architecture supporting up to 1M tokens at ~9x prefill / 15x decode speedup over M2 at 1/20th the per-token compute. It scores 59.0% on SWE-Bench Pro (surpassing GPT-5.5 and Gemini 3.1 Pro), 83.5 on BrowseComp (vs Opus 4.7's 79.3), 66.0% Terminal Bench 2.1, and 74.2% MCP Atlas. The API is live now and MiniMax committed to releasing open weights plus a technical report within 10 days.