The short version
Sparse-attention model chasing million-token context on less compute.
From the HackerLinks archive
SubQ 1.1 Small
Sparse-attention model chasing million-token context on less compute.
At a glance:
First seen:2026-06-16
Last seen:2026-06-16
Times seen:1
Website:subq.ai
Why it caught our attention
Commenters praised the tech and highlighted its large compute and speed gains.
Where it surfaced on Hacker News
2026-06-16
Editorial paraphrase
One commenter said they loved the tech; another pointed to the 64.5x compute reduction and 56x speedup claim.
Original threadSubQ 1.1 Small