Rumble-2-Flash released: a 0.5B speech recognition model supporting more than 40 languages

Key Details Reson8 released Rumble-2-Flash, a 0.5B-parameter speech recognition model, under Apache 2.0. It provides automatic language detection and output controls for numbers, filler words, and repeated expressions, and can run in Transformers and vLLM without custom user code. Use Cases The supplied Transformers example takes audio input, constructs a transcription request, and extracts the transcript from the model output. Audio inputs can be local paths, URLs, NumPy arrays, or lists of these. Costs and Limitations The weights are licensed under Apache 2.0. A separate API is also available, but pricing is unconfirmed. The reported throughput of over 7,000 RTFx was measured on a large evaluation dataset with sufficiently full batches.

Visit original source

We remember your site preferences. If you allow analytics, we also measure visits and page activity.

Analytics is off.

Cookie policy