255 Large-Scale AI Speech / Audio Model Competition = +367% Increase in Models Released Y/Y, per Epoch AI Note: An AI system can operate in more than one domain and may be double-counted across pages. Includes models without verified training compute. Many models shown are multimodal. Source: Epoch AI (5/25), Microsoft (1/23), OpenAI (5/24), Amazon, Pinterest Speech / Audio Models – Examples Speech / Audio Models – Releases 0 5 10 15 2017 2018 2019 2020 2021 2022 2023 2024 +367% Rising Competition = AI Model Releases Number of Systems Released per Year OpenAI GPT 4o Speech – 5/24 Microsoft VALL-E – 1/23