The update adds automatic speech recognition for three models and support for the NVIDIA B200 GPU. The company says it outperformed each model’s reference engine in comparable tests.
Photon 2.1 adds streaming automatic speech recognition for Whisper, Qwen3-ASR, and Parakeet, as well as support for the NVIDIA B200. Moondream says the system outperformed each model’s reference engine in every comparable test; this is the company’s claim, not an independent assessment provided in the briefing.
For teams evaluating self-hosted speech recognition, the report can be a starting point for comparing performance across models and GPU types. When using AI to study or apply these results, avoid sending identifiable recordings without authorization and remove or replace personal data where possible.