# Amazon releases Mitra-v2 tabular foundation models under Apache-2.0

> Original title: Mitra-v2: Scaling synthetic pretraining for tabular foundation models

- Source: Amazon Science
- Published: 2026-11-10T15:37:39.000Z
- Bad Influence: https://badinfluence.news/items/iqlfgr116w2mh1fmrqtlh2gre
- Original: https://www.amazon.science/blog/mitra-v2-scaling-synthetic-pretraining-for-tabular-foundation-models

## Summary

Amazon releases Mitra-v2, a second-generation tabular foundation model for classification and regression, on Hugging Face under the Apache-2.0 license. The model is pretrained on synthetic tables with 10 times as many rows and roughly three times as many columns as Mitra-v1, and it ranks first among single models on TabArena. Amazon Connect Customer, which led the effort, is adopting it for predictions on structured business and customer-experience data.

## Why it matters

The post explains how synthetic pretraining and wider table sizes changed the model's limits, which helps readers judge tabular foundation models for their own data.