Google Unveils Muir Linear Attention Architecture

Google Research released Muir, a linear attention architecture built on Gated DeltaNet-2 that streams 10 million tokens with only 25MB of state and trains 2.7x faster, while improving long-context recall.

2026-11-07 ~ 2026-11-07 · 3 related posts

• Google Proposes Muir, a Linear Attention Architecture Beating Gated DeltaNet-2 and Transformers — google · 2026-11-07
• Google's Muir linear attention fits 10M tokens in a 25MB state, beats GDN-2 — CategoryNormal149 · 2026-11-07
• Muir linear attention streams 10M tokens in 25MB state, trains 2.7x faster than Transformer at 32K — _akhaliq · 2026-11-07