Sources used for the initial model database
We prefer official model cards and repositories. External benchmark articles can be useful context, but they are not treated as the authoritative source for supported languages or licenses.
OpenAI WhisperOfficial GitHub repository ↗
Model sizes, approximate VRAM guidance, turbo description, multilingual behavior and MIT license.
NVIDIA Parakeet TDT 0.6B v3Official model card ↗
600M model, 25 listed languages including Slovenian, CC BY 4.0, current downloadable formats.
NVIDIA Nemotron 3.5 ASR Streaming 0.6BOfficial model card ↗
Native streaming, tiered language-locales, Slovenian adaptation-ready status and OpenMDW-1.1 license.
Qwen3-ASR 0.6B and 1.7BOfficial GitHub repository ↗
Released model sizes, official language list, offline/streaming inference and timestamp tooling. Model cards list Apache-2.0.
Distil-Whisper large-v3Official model card ↗
English-only Distil-Whisper large-v3 and MIT license.
Moonshine AIOfficial repository ↗
English model sizes, streaming variants and model-specific licensing. English models are MIT; non-English models can use different terms.
Corrections
If a primary source has changed, contact contact@speechmodelmatch.com with the exact model and source link.