Berlin-based startup Deepslate has successfully closed a seed funding round, securing €7.7 million. The investment was led by Munich-based 42CAP, with additional backing from Alstin Capital, SIVentures, and various business angels.
The company specializes in speech-to-speech AI models that process audio directly, bypassing traditional text-conversion steps. This architecture utilizes a proprietary speech encoder, a reasoning core based on post-trained open-weights models, and a speech decoder.
According to an independent Artificial Analysis benchmark from September 2026, the system achieved a response time of 440 milliseconds, ranking as the fastest model tested. Furthermore, the technology holds the top error rate performance for European languages in the CoVoST2 benchmark.
Deepslate maintains ISO 27001 certification and hosts its infrastructure entirely within the European Union. The platform is currently deployed by contact centers, insurers, and various digital platforms via API and self-service options.
The newly acquired capital will support the expansion of the company's data program, specifically targeting German dialects, personal names, and street names. Additionally, Deepslate plans to scale its production infrastructure and grow its sales and marketing teams.
