Stability AI models, access and benchmark guide
Stability AI model and access guide with a practical US-English evaluation brief and links to publisher materials.

👁 111 reads
Stable Audio spans hosted generation and downloadable models. The cost of a local studio setup comes from hardware and license eligibility, while a hosted API can bill a defined number of credits per result. Put the checkpoint size on both listening notes and estimates.
Choosing the first job
Layer an original ticking rhythm beneath narration and check for unexpected music across a planned silent beat.
Models in the November 17 snapshot
| Model and track | Placement | Reported score |
|---|---|---|
| Stable Audio 3 Large (AA instrumentals) | 8 | 1014 Elo |
| Stable Audio 3 Medium (AA instrumentals) | 14 | 996 Elo |
| Stable Audio 2.5 (AA instrumentals) | 16 | 984 Elo |
| Stable Audio 3 Small (AA instrumentals) | 17 | 946 Elo |
Release record. Publisher documentation identifies the product family. The ranked configurations include Stable Audio 3 Large, Stable Audio 3 Medium, Stable Audio 2.5.
Google · Your sourcesQuantslant higher in your Google - one clickAdd to preferred sources →Sources
- https://stability.ai/stable-audio (accessed November 17, 2026)
- https://huggingface.co/stabilityai (accessed November 17, 2026)
- https://artificialanalysis.ai/embed/music-generation-leaderboard/leaderboard/instrumental (accessed November 17, 2026)
More on this from QUANTSLANT
benchmarks·SpeechifyAI models, access and benchmark guideSpeechifyAI model and access guide with a practical US-English evaluation brief and links to publisher materials.