Zyphra scales subquadratic attention ZHA to trillion-param open models with 9-60x long-context speedups

realmrfakename · x · 2026-11-18

Zyphra announced it has scaled its subquadratic attention method, ZHA, to leading open-source multi-trillion-parameter models including Nemotron, Qwen, Mimo, and GPT-OSS. The company claims ZHA preserves reasoning and agentic performance while delivering 9-60x long-context speedups at 10M tokens.

Original post →

More from Infra

Infra channel →