Reson8 releases Rumble-2-Flash, a compact 0.5B speech recognition model
In short: Reson8 has launched Rumble-2-Flash, an open-source 0.5B parameter speech-to-text model supporting over 40 languages. It achieves performance parity with models up to 30B parameters on noisy, real-world call-center benchmarks.
This summary was generated automatically by AI from Hugging Face's publication. It is our own text, not a copy of the original — facts, figures and quotes belong to the source, linked above and below.
What changed?
- 1Released under Apache 2.0 license
- 2Achieved 7.86% WER on call-center benchmarks, matching 30B models
- 3Native support in Transformers and vLLM
- 4Throughput of over 7,000 RTFx
| Parameter | Before | Now |
|---|---|---|
| Parameter count | Not specified | 0.5B |
| Throughput | Not specified | 7,000+ RTFx |
Why it matters
This model drastically lowers the hardware requirements for high-accuracy voice transcription in production contact-center environments, allowing smaller footprint deployments without sacrificing reliability on compressed, noisy phone audio.
What it means for AI agents and contact centers
Integrate Rumble-2-Flash into your pipeline via standard Transformers or vLLM to significantly reduce latency and compute costs for voice agents. Compare this 0.5B model against your current transcription providers, especially focusing on WER in noisy, real-world call environments.
🧪 Worth Testing
The extreme efficiency and high performance on phone-call benchmarks make this an ideal candidate to reduce the infrastructure overhead of real-time voice AI agents.
Sources
- Hugging FaceOfficialPrimary sourceOriginal article →„Rumble-2-Flash: an open 0.5B speech model for 40+ languages“18 Nov 2026, 02:00
- Published by source
- 18 Nov 2026, 02:00
- Found by our system
- 18 Nov 2026, 16:56
- Summary generated
- 18 Nov 2026, 18:26
This article was written by AI from the original source. Facts, numbers and prices come from the source; missing values are marked “Not specified”. Legal notice, copyright and privacy