Published October 2, 2026
| Version 0.5.3
Software
Open
Kavier: Simulating the Performance, Sustainability, and Efficiency of LLM Ecosystems under Inference and Training
Authors/Creators
- 1. Vrije Universiteit Amsterdam
- 2. IBM Research
Contributors
Research group:
Description
LLM workloads, such as training, fine-tuning, and inference, are becoming too large even for massive-scale infrastructure. Kavier is a physics-driven simulator that predicts the performance, sustainability, and efficiency of LLM ecosystems under inference and fine-tuning workloads.
Files
atlarge-research/Kavier-v0.5.3.zip
Files
(437.0 kB)
| Name | Size | Download all |
|---|---|---|
|
md5:fdc6e76a35a2e67f651a37cdc0d70704
|
437.0 kB | Preview Download |
Additional details
Related works
- Is described by
- https://atlarge-research.com/pdfs/2025-15-07_bsc_thesis_radu_nicolae.pdf (URL)
- Is documented by
- https://atlarge-research.github.io/Kavier/ (URL)
Software
- Repository URL
- https://github.com/atlarge-research/Kavier