Latest works

Tarka Embedding 30M V1

Release | 7-1-2026

We achieved 20× compression and recovered ~86% MTEB performance. The model supports elastic upscaling at inference, avoiding full-weight loading and reducing both memory footprint and compute cost.

Open - Source

Reduce and Refine

Release | 1-12-2025

This work explores model compression by progressively reducing a 28-layer model to a lean 6-layer model without major performance loss. Reduce and Refine demonstrates a practical path to faster, lighter, more efficient LLMs.

Open - Source

Tarka Embedding V1

Release | 9-11-2025

The Tarka Embedding V1 series is a compact and efficient text embedding model family developed to explore the capabilities of knowledge distillation, coreset selection, and model compression techniques

Open - Source