Frontend Engineer

Deepslate lands €77M for speech models that never leave the EU. Deepslate's end-to-end speech models cut the text round trip and keep European voice data inside EU data centers. Last updated: October 3, 2026 12:17 am Deepslate's bet is that European voice AI can stay inside the bloc without losing speed, and the Berlin company now has €77M to prove it. Conventional stacks pass speech through three stages: audio becomes text, a language model composes a reply, and the reply is turned back into audio. Deepslate throws out that relay, running its own speech-to-speech models that take audio in and hand audio back, holding onto tone, emphasis and dialect while shaving latency. Three in-house-trained pieces make up the system: an encoder for speech, a reasoning core that rests on an open-weights language model Deepslate post-trains language by language, and a decoder that speaks the answer. Swapping the model underneath is possible without rebuilding everything around it. The company points to numbers. In an independent Artificial Analysis benchmark its model posted a 440-millisecond response time, which Deepslate calls the quickest speech-to-speech system measured as of September 2026, and it claims the best error rate among European languages in CoVoST2. It also holds ISO 27001 certification. Revenue is already coming in, with insurers, contact centers and platforms running the model in production, bought through a self-service platform and API or self-hosted at higher volume. Sovereignty is the selling point: Deepslate argues its customers treat it as a prerequisite rather than a bonus, and that it counts for nothing unless it can be verified. The new money goes to model training and European data work, particularly German street names, personal names and dialects, plus sales hiring and added capacity in EU data centers.

Deepslate secures €7.7M to expand its European voice AI platform. Published: Oct 1, 2026, 07:21 AM EDT Updated: Oct 1, 2026, 07:21 AM EDT Berlin-based voice AI company Deepslate has raised €7.7 million in seed funding to accelerate the development and expansion of its voice AI technology. The round was led by Munich-based technology investor 42CAP, with participation from Alstin Capital, existing investor SIVentures, and several business angels. Deepslate develops its own speech-to-speech AI models that process spoken language directly instead of first converting speech into text. The company focuses on European languages and operates its technology entirely within the European Union. Traditional voice AI systems typically follow several steps: speech is converted into text, the text is processed by a language model, and the response is then converted back into speech. Deepslate takes a different approach with an end-to-end model that processes incoming audio directly and produces an audio response. The company says this can reduce response times while preserving details of spoken communication, including tone, emphasis, dialect, and other vocal characteristics. Its technology is built around three components developed and trained in-house: a speech encoder, a reasoning engine based on an open-weights language model that Deepslate post-trains for individual languages, and a speech decoder that turns the response into spoken audio. This architecture also allows the company to replace the underlying language model without having to retrain the entire system. According to Deepslate, its model achieved a response time of 440 milliseconds in an independent Artificial Analysis benchmark, making it the fastest speech-to-speech model measured in that benchmark as of September 2026. The company also reports that its technology delivered the lowest error rate in the comparison for European languages in the CoVoST2 benchmark. Deepslate is also ISO 27001 certified. The technology is already being used in production by insurers, contact centres, and other platforms. Deepslate provides access to its models through a self-service platform and API, while platform providers and enterprise customers can also access volume and self-hosting options. For its customers, data sovereignty is not a nice-to-have, it is a prerequisite. And either it can be verified or it is worthless. That is why Imperium Magazine disclose where the computing happens, who its subprocessors are and what is still open, said Paskal Paesler, co-founder of Deepslate. With the new investment, Deepslate plans to expand its model training and European data initiatives, with a particular focus on German street names, personal names, and regional dialects. The company will also work on reducing response times and improving overall voice quality. At the same time, Deepslate plans to strengthen its sales and marketing teams and expand its production infrastructure across European data centres to handle growing call volumes.