TLDR 2026-10-01

Modulate raised $25M, led by Future Ventures, to build the understanding layer for voice AI.

Transcripts drop the fraud, frustration, and abuse hiding in conversations. Velma, Modulate's audio-native model, catches them by orchestrating 100+ voice-specialized submodels running on raw audio.

✅ Stop fraud→ catch deepfakes with the #1-ranked detection model on Hugging Face✅ Catch AI agent failures→ flag hallucinations and compliance violations in real time✅ Keep customers→ spot churn and dissatisfaction signals✅ Protect human and AI agents→ flag risk and abuse✅ Build your foundation with it→self-serve APIs; top ranked Hugging Face Leaderboard STT API

The funding brings more models, deployment options, and broadens the APIs for teams building with voice.

Start building with Modulate's API

Google's Gemini 4 Argon model offers industry-leading performance in coding, knowledge work, and cybersecurity. Google's engineers are already using the model extensively, and Argon-powered agents have been working to migrate C/C++ codebases to Rust across Google. The model supports an output limit of 1 million tokens. It will be priced at $2 per million input tokens and $10 per million output tokens for a limited time. The model is still in limited testing, and there is no timeline for when it will be available to enterprise and consumer customers.