76.0
▲ 30% since 2019
top average score on MTEB (English v2) in 2025, up from 64.8 in 2023
20192025
What the chart shows
The Massive Text Embedding Benchmark tests the models that turn text into vectors for search, clustering and retrieval, across more than 50 datasets in eight task types. The best average score on its English v2 set reached 76.0 in 2025, up from 70.7 in 2024 and 64.8 in 2023.
Progress was slow at first: the top score rose by less than one point between 2019 (58.3) and 2021 (59.0). The best models remain well short of a perfect 100.
Key numbers
| Measure | Value |
|---|---|
| 2025 | 75.97 |
| 2024 | 70.72 |
| 2023 | 64.77 |
| 2022 | 63.13 |
| 2021 | 59.03 |
| 2019 | 58.27 |
How to read it
- Each bar is the highest average task score among models on the leaderboard for that year.
- The data has no value for 2020.
- The average combines different metrics across tasks (for example accuracy for classification and ranking quality for retrieval) on a 0–100 scale.
Source: MTEB Leaderboard, 2026, via the AI Index Report 2026 (Stanford HAI), figure 2.2.4: MTEB (English v2): average score. Chart and text by AI Adoption.
Numbers checked against the published data on 20 October 2026.
Download the data (CSV) · Cite this chart
Numbers checked against the published data on 20 October 2026.
Download the data (CSV) · Cite this chart
Discussion