Amazon's Mitra-v2: synthetic-data tabular foundation model matches 1.6B rivals with 77M params
AxSaucedo · x · 2026-11-20
Amazon released Mitra-v2, a tabular foundation model pretrained purely on synthetic data from a hybrid causal-graph prior. With a small 2D Transformer backbone (77M params), it hits SOTA on TabArena, rivals industry-scale TabFM and EXAONE while beating TabPFN-3, and matches a 1.6B-param model with 5% of its size. Weights, inference and fine-tuning code are released under Apache-2.0.
More from Models
- Even with OpenAI's Daybreak access, Codex still blocks cybersecurity requests — AIFlow_ML · 2026-11-20
- DeepSeek v4.1 flash reportedly served at higher quality for 10% of the price — _xjdr · 2026-11-20
- Olala mini: 7B total / 1B active model trained on 13T tokens mixes Mamba-3 with differential attention — LucaAmb · 2026-11-20
- Open-source decision model Laya hits 35k stars, outputs calibrated probabilities in 33ms — kingslyj · 2026-11-20
- Google's Logan Kilpatrick teases Gemini subscription changes starting December 2 — teortaxesTex · 2026-11-20
- Analyst: Opus 5.5 jump wasn't compute-limited — China's real gap is data engineering — teortaxesTex · 2026-11-20