https://github.com/moiri-gamboni/embedding-bench
Interactive explorer for sub-1B open embedding models across MTEB retrieval benchmarks (NDCG@10)
https://github.com/moiri-gamboni/embedding-bench
benchmark embedding embedding-models mteb retrieval
Last synced: 25 days ago
JSON representation
Interactive explorer for sub-1B open embedding models across MTEB retrieval benchmarks (NDCG@10)
- Host: GitHub
- URL: https://github.com/moiri-gamboni/embedding-bench
- Owner: moiri-gamboni
- Created: 2026-02-11T10:15:26.000Z (5 months ago)
- Default Branch: main
- Last Pushed: 2026-06-15T11:51:35.000Z (about 1 month ago)
- Last Synced: 2026-06-15T13:23:42.235Z (about 1 month ago)
- Topics: benchmark, embedding, embedding-models, mteb, retrieval
- Language: HTML
- Size: 456 KB
- Stars: 0
- Watchers: 0
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# Embedding Model Benchmark Explorer
Interactive browser-based tool for comparing sub-1B parameter open embedding models across multiple retrieval benchmarks.
## Benchmarks
- **Prose** -- general text retrieval (ArguAna, FiQA, TRECCOVID, etc.)
- **RTEB Code** -- code retrieval (HumanEval, MBPP, DS1000, etc.)
- **CoIR** -- code information retrieval (CodeSearchNet, CosQA, StackOverflowQA, etc.)
- **Instruct** -- instruction-following retrieval
- **LongCtx** -- long context retrieval (LEMB suite)
- **Reason** -- temporal reasoning retrieval
All scores are NDCG@10.
## Usage
Open `index.html` in a browser. The data is bundled in `models.js` so it works from `file://` with no server.
## Updating the data
Pull fresh results from the [MTEB results repo](https://github.com/embeddings-benchmark/results):
```bash
git clone --depth 1 https://github.com/embeddings-benchmark/results /tmp/mteb-results
python pull_results.py
```
Optionally merge RTEB Code scores from a HuggingFace leaderboard CSV export:
```bash
python pull_results.py --csv ~/Downloads/rteb-code-leaderboard.csv
```