Kimi Linear: An Expressive, Efficient Attention Architecture
Kimi Team et al.
arXiv:2510.26692
In short
Kimi Linear is a hybrid architecture whose linear-attention layers keep a fixed-size memory instead of a growing cache, mixed with a few full-attention layers. With the same training recipe it outperforms full attention while cutting cache memory by up to 75% and speeding decoding up to 6× at a million tokens.
Why it matters
It is a credible claim that linear attention can replace full attention without a quality penalty.
Read first
The 4 Field Guide ideas this paper leans on.
Starting from scratch? The full route 16 ideas · basics first
- Machine Learning ✓ understood
Building systems that learn patterns from data instead of following hand-written rules, getting better at a task as they see more examples.
- Neural Network ✓ understood
A model built from layers of simple units, each taking a weighted sum of its inputs and applying a nonlinearity; training adjusts the weights so the whole stack maps inputs to useful outputs.
- Dataset ✓ understood
A collection of data examples used for training, validating, or testing machine learning models.
- Feature ✓ understood
A single measurable property of an example, such as a house's floor area or how many links an email contains, used as an input to a model.
- Deep Learning ✓ understood
A subset of machine learning that uses neural networks with multiple layers (deep neural networks) to learn hierarchical representations of data.
- Representation Learning ✓ understood
Letting a model learn its own internal description of the data, such as embeddings or layered visual features, instead of relying on features designed by hand.
- Embedding ✓ understood
A list of numbers (a vector) that represents a word, sentence, image or other item, learned so that similar items end up close together.
- Attention Mechanism ✓ understood
A technique that lets a neural network weigh every part of its input when producing each output, focusing on the parts most relevant at that step.
- Self-Attention · read first ✓ understood
An attention step in which every token in a sequence looks at every token in that same sequence and builds a new representation from a weighted mix of them.
- Recurrent Neural Network · read first ✓ understood
A neural network architecture with loops that allow information to persist, designed for sequential data like text and time series.
- Token ✓ understood
The unit of text a language model reads and writes: a word, part of a word, or a character, drawn from the model's fixed vocabulary.
- Transformer ✓ understood
A neural network architecture, introduced in 2017, built from stacked self-attention and feed-forward layers; the basis of nearly every modern large language model.
- Context Window · read first ✓ understood
The maximum number of tokens a language model can take into account at once, counting both the input it reads and the output it writes. Also called context length.
- Training ✓ understood
The process of fitting a model to data by repeatedly measuring how wrong its outputs are and adjusting its parameters to reduce that error.
- Inference ✓ understood
Running a trained model on new inputs to get predictions, with its weights frozen: the stage of a model's life that users actually interact with.
- Throughput · read first ✓ understood
The number of predictions or tokens a model can process per unit of time, a key deployment performance metric.
In the frontier
- Rank
- #42 of 100
- Citations
- 116
- as of Aug 9, 2026
- Published
- Oct 2025
Topics: Model architecture , Long context and attention , Efficiency and serving
Selection: 1kpapers.com by Together AI, most-cited as of Aug 9, 2026