MiniMax's natively multimodal MoE (428B total / 23B active) with a new sparse-attention architecture and 1M context — open-sourced June 2026 and ranked the top open-weight model on the Artificial Analysis Intelligence Index, beating GPT-5.5 and Gemini 3.1 Pro on SWE-Bench Pro.
MiniMax's natively multimodal MoE (428B total / 23B active) with a new sparse-attention architecture and 1M context — open-sourced June 2026 and ranked the top open-weight model on the Artificial Analysis Intelligence Index, beating GPT-5.5 and Gemini 3.1 Pro on SWE-Bench Pro.
Parameters, context, and license as reported June 2026. Confirm the current license on the official Hugging Face page before deployment.
← Back to all models