Calibration-free compression brings Evo 2 to its full million-token context on a single GPU
- 1The University of Texas at Austin
Abstract
Evo 2 is the largest openly available genomic foundation model, but its forty billion parameter configuration cannot be loaded onto a single 80 GB accelerator, placing genome-scale analysis beyond most laboratories. We present TurboQuant-Bio, an open toolkit that compresses Evo 2s weights and attention cache to four bits without calibration data, and serves both through fused kernels. Compression is near-lossless across perplexity spanning the tree of life, genomic classification, splice-site prediction, gene completion and clinically relevant variant-effect prediction. It brings Evo 2 40B onto one 80 GB GPU and Evo 2 7B to its full million-token context within a 40 GB memory budget, an eightfold gain in reachable context. We further show that the released chunked-prefill path is silently incorrect, returning plausible but uncorrelated likelihoods, and derive the block-wise continuation that repairs it: a complete 580-kilobase bacterial genome is now scored in one context in 22 minutes rather than 13.7 hours.
Related
- Relational Knowledge Distillation Brings DNN Representations Close Enough to Humans to Be Aligned Without Supervision
- Beyond a single threshold: calibration of CADD v1.7 and AlphaGenome for ACMG/AMP computational evidence across variant classes, model versions, and reference data
- Harmonised benchmarking of foundation models for single-cell and spatial transcriptomics reveals context-dependent generalisation
- Efficient genome-wide mapping of reproducible, context-dependent eQTLs at single-cell resolution
- LoCoFlow: Local-Context Flow Matching for Single-Cell Dynamics and Fate Inference
BibTeX
@misc{patsakis2026calibrationfree,
title = {{Calibration-free compression brings Evo 2 to its full million-token context on a single GPU}},
author = {M. Patsakis and A. Tzanakakis and I. Georgakopoulos-Soares},
howpublished = {bioRxiv},
year = {2026},
month = aug,
doi = {10.64898/2026.08.28.747902},
url = {https://www.biorxiv.org/content/10.64898/2026.08.28.747902v1},
note = {CC BY},
}
RIS
TY - UNPB
TI - Calibration-free compression brings Evo 2 to its full million-token context on a single GPU
AU - M. Patsakis
AU - A. Tzanakakis
AU - I. Georgakopoulos-Soares
PY - 2026
DA - 2026/08/28
PB - bioRxiv
DO - 10.64898/2026.08.28.747902
UR - https://www.biorxiv.org/content/10.64898/2026.08.28.747902v1
AB - Evo 2 is the largest openly available genomic foundation model, but its forty billion parameter configuration cannot be loaded onto a single 80 GB accelerator, placing genome-scale analysis beyond most laboratories. We present TurboQuant-Bio, an open toolkit that compresses Evo 2s weights and attention cache to four bits without calibration data, and serves both through fused kernels. Compression is near-lossless across perplexity spanning the tree of life, genomic classification, splice-site prediction, gene completion and clinically relevant variant-effect prediction. It brings Evo 2 40B onto one 80 GB GPU and Evo 2 7B to its full million-token context within a 40 GB memory budget, an eightfold gain in reachable context. We further show that the released chunked-prefill path is silently incorrect, returning plausible but uncorrelated likelihoods, and derive the block-wise continuation that repairs it: a complete 580-kilobase bacterial genome is now scored in one context in 22 minutes rather than 13.7 hours.
ER -