Resonate Labs B.V., the Amsterdam speech recognition startup known as Reson8, released its Rumble-2-Flash transcription model on Wednesday as a free download with open weights. The 500 million-parameter model is built for fast transcription of prerecorded audio and supports more than 40 languages. Reson8 is aiming the release at developers and researchers who need speech recognition for languages other than English, a group the company argues has been underserved by the leading speech models.

According to Reson8, the top speech models are trained mainly on English even though roughly one in 20 people speaks it as a first language. The company says accuracy drops furthest on specialist terms, such as medication names, once speech moves outside English. Reson8's models are in production use for meeting notetakers, artificial intelligence receptionists, customer support voice agents, medical transcription and captioning, according to the company.

Co-founder and Chief Executive Raoul Ritter said generic speech models are built for English and make noticeably more mistakes in other languages. Ritter, who started Reson8 with Thomas Kluiters and Jarno Verhagen, said languages spoken by millions, including Dutch, Swedish and German, are treated as an afterthought. Kluiters, the company's head of AI, said the models are trained for noisy settings such as phone-line audio, adding that the real world does not happen in a recording studio.

Reson8 describes Rumble-2-Flash as one of the smallest models on Hugging Face Inc.'s Open ASR Leaderboard. In the company's own runs of the leaderboard's multilingual tests, the model averaged a word error rate of 7.09% across seven languages, including German, French and Dutch. Reson8 claims an average error rate of about 5% in English. On a separate test covering 10 lower-resource languages such as Maltese and Basque, the company says Rumble-2-Flash posted a mean word error rate of 17.6%, compared with 39.82% for OpenAI Group PBC's Whisper Large V3 on the same test.

The open-weights release follows the model's Nov. 3 debut in Reson8's application programming interface and comes after the company launched its flagship Rumble-2 model. Custom vocabulary support, which lets customers add up to 50,000 terms as text without fine-tuning, remains exclusive to Rumble-2. Rumble-2-Flash is available free under an Apache 2.0 license and can also be accessed through Reson8's API at prices starting from €0.01 ($0.001) per hour of audio.

Reson8 says its service runs on the company's own graphics processing units in the Netherlands and retains no audio or transcripts. The company also claims ISO 27001 certification and compliance with the European Union's General Data Protection Regulation and the bloc's AI Act. Reson8 raised €5 million ($5.5 million) in pre-seed funding in March, in a round led by Balderton Capital with NP-Hard Ventures participating. James Wise, a partner at Balderton, said speech is quickly becoming a default interface across software.

More software news from TechManNews.