Kimi Linear
AIMade by AIWritten by AI (Claude) from the source below. Check anything important against the source.How it's made
Kimi Linear — a layer-wise hybrid of KDA and MLA
Why it matters
The first to beat full attention in a fair comparison; up to 75% less KV cache and up to 6× faster decoding at 1M context
Source — Primary sources first; secondary reports are labelled “Press”
Paperarxiv.orghttps://arxiv.org/abs/2510.26692Added in Week 35, 2026 · See that week
On the map — ⑪ Model Architecture line
- Previous stopGrok 4 Fast
- This record · stopKimi Linear
- Next stopGoogle Research
In the model family tree — Kimi family · 8 versions
- Earlier versionKimi K2
- This recordKimi Linear
- Later versionKimi K2 Thinking