long-context

taxonomy axis: design_intent

Decision rule: Framed primarily around context length.

Anchor: Gemini 1.5 (if in atlas)

Models in this category (3)

ModelOrgDateBlockAttentionContextTotalActive
LongCat-FlashMeituan LongCat Team2026-01-29sparse-MoEnot disclosed131,07268.5B2.9B
DeepSeek-V4DeepSeek-AI2026-04-26sparse-MoEhybrid1M1.6T49B
MiniMax M3MiniMax (with authors from Peking University, NVIDIA, Zhejiang University, HUST, Nanjing University, Hangzhou Dianzi University)2026-06-11sparse-MoEGQAnot disclosed109B6B

What varies within this category

attention variant varies — GQA, hybrid, not disclosed. trained context varies — 1000000, 131072, not disclosed. scale class varies — frontier, large.