Microsoft Research publishes the Phi family of small open-weight models that punch above their parameter count on reasoning benchmarks. Phi models are designed for edge and on-device inference.
Visit MicrosoftModels tracked
4
Open weight
4
API only
0
Avg score
66.7
Top benchmark
80.0
STS
Total HF downloads
789.1K
First release
Feb 2025
Latest release
Mar 2026
Ranked by composite score across benchmarks, popularity, efficiency, and versatility.
| # | Model | Modality | Score | Params | Released |
|---|---|---|---|---|---|
| 1 | embedding | AA72.5 | — | Mar 2026 | |
| 2 | embedding | AA70.3 | — | Mar 2026 | |
| 3 | audio | BB64.9 | 5.6B | Feb 2025 | |
| 4 | embedding | BB59.0 | 27B | Mar 2026 |
The strongest model from this provider in each modality. Click a card to open the model page.
How this provider’s catalog splits across text, image, video, audio, and embedding.
When each model shipped, newest first.
Mar 30
Mar 30
Mar 30
Composite grades across this provider’s catalog. Higher is better, blending benchmarks, popularity, and efficiency.
Models with downloadable weights, ranked by composite score.
| # | Model | Modality | Score | Params | Released |
|---|---|---|---|---|---|
| 1 | embedding | AA72.5 | — | Mar 2026 | |
| 2 | embedding | AA70.3 | — | Mar 2026 | |
| 3 | audio | BB64.9 | 5.6B | Feb 2025 | |
| 4 | embedding | BB59.0 | 27B | Mar 2026 |
Each family is a model series. Browse all releases in a family with their own leaderboard.
Full Directory
Open the full directory to filter by hardware, capability, license, and benchmark score.
Or Browse by Family
See every release in a family side by side, with a timeline and a leaderboard. Useful when you have already picked a series and want the right size.