Heimr 570M: a local model that stays fast as the context grows | Hacker News - But Better

Heimr 570M: a local model that stays fast as the context grows sevren.ai

Heimr 570M is a 2.7B-parameter, 570M-active hybrid of Mamba-3 and Heimr attention with mixture-of-experts layers, built for phones and laptops. Its decode throughput and memory stay nearly flat to 64K tokens of…

1 points•ilreb•about 12 hours ago•0 comments•

#heimr #570m #heimr 570m #context #local #local model #context grows #stays fast

0 comments

No comments yet.

Related stories

Show HN: Ox – A local agent that uses the internet for youopenox.ai
Hacker News · 2 points · 7 days ago
Show HN: Decide – Jev decisions in the shell, scripts, and agent skillsgithub.com
Hacker News · 1 points · 2 days ago
Show HN: Local-coder – Build a team of coding agents with local modelsgithub.com
Hacker News · 6 points · 10 days ago
Rosenbjerg/local-review: Minimal webapp for local code review of branchesgithub.com
Hacker News · 3 points · 10 days ago
Cities across US oppose Trump FCC plan to preempt local broadband rulesarstechnica.com
Ars Technica · 0 points · 8 days ago
Open Source Polygo: translate your app's strings with a local model, Git-nativegithub.com
Hacker News · 1 points · 9 days ago