Introducing LongCat-2.0 ๐ฑ
1.6T parameters ยท MoE with ~48B active ยท 1M context
The full model behind Owl Alpha on @OpenRouter โ now available.
Built for agentic coding from the ground up:
โ LongCat Sparse Attention (LSA) โ scales efficiently for 1M-context tokens
โ Zero-Compute Experts โ dynamic activation 33Bโ56B per token, zero wasted compute
โ MOPD โ three specialized expert groups (Agent / Reasoning / Interaction), gate-routed per task
How it stacks up:
โ Terminal-Bench 2.1: 70.8
โ SWE-bench Pro: 59.5 (GPT-5.5: 58.6)
โ SWE-bench Multilingual: 77.3
โ FORTE: 73.2 ยท RWSearch: 78.8 ยท BrowseComp: 79.9
๐ Tech Blog: https://t.co/4KrjyKiDBn
Try it across different scenarios ๐งต๐
semoga dengan kejadian kemarin, membuat dek @primagnesius lebih humble, berfikir dampak dari tulisan maupun perkataan, instrospeksi.
Notes : bukan saya yang chat pak sabirin