# Zeta2 Improves Zed Edit Predictions with Larger Training Data and New Context Retrieval

DevFeed: [Zeta2 Improves Zed Edit Predictions with Larger Training Data and New Context Retrieval](<https://devfeed.tech/articles/we-rebuilt-zeta-from-the-training-data-up-13599.md>)

Original publisher: [Read original article](<https://zed.dev/blog/zeta2>)

Author: Ben Kunkle, Max Brunsfeld, Oleksiy Syvokon

Published: 2026-03-25T00:00:00Z

Content type: article

Language: en

Sources: [Zed Industries - Blog](<https://devfeed.tech/sources/zed-industries-blog.md>)

Topics: [Training AI Models](<https://devfeed.tech/topics/training-ai-models.md>), [Open Source Models & Datasets](<https://devfeed.tech/topics/open-source-models-datasets.md>), [datasets](<https://devfeed.tech/topics/datasets.md>), [model-deployment](<https://devfeed.tech/topics/model-deployment.md>), [Latency](<https://devfeed.tech/topics/latency.md>), [Code](<https://devfeed.tech/topics/code.md>)

Tags: [code](<https://devfeed.tech/tags/code.md>), [evaluation](<https://devfeed.tech/tags/evaluation.md>), [latency](<https://devfeed.tech/tags/latency.md>), [model](<https://devfeed.tech/tags/model.md>), [open-source](<https://devfeed.tech/tags/open-source.md>), [training](<https://devfeed.tech/tags/training.md>)

## AI overview

Zed describes Zeta2, its updated edit prediction model, which has a 30% higher acceptance rate than Zeta1 and is now the default for Zed users. The model was trained on nearly 100,000 opt-in examples from open-source repositories, uses improved context retrieval, and is open-weight. The article also discusses changes to training, evaluation, deployment, and feedback infrastructure.

## Source excerpt

Zeta2 is our latest edit prediction model and it's 30% better than Zeta1.