Audio Flamingo 3: Advancing Audio Intelligence with Fully Open Large Audio Language Models

TL;DR

Audio Flamingo 3 is a fully open large audio-language model that significantly advances reasoning and understanding across speech, sound, and music through novel training strategies and multi-modal capabilities.

Contribution

It introduces AF-Whisper, a unified audio encoder, and a comprehensive training curriculum, enabling state-of-the-art performance on diverse audio understanding benchmarks.

Findings

Achieves new SOTA on over 20 audio benchmarks.

Trained solely on open-source data, surpassing larger closed models.

Supports long audio reasoning up to 10 minutes and voice-to-voice interaction.

Abstract

We present Audio Flamingo 3 (AF3), a fully open state-of-the-art (SOTA) large audio-language model that advances reasoning and understanding across speech, sound, and music. AF3 introduces: (i) AF-Whisper, a unified audio encoder trained using a novel strategy for joint representation learning across all 3 modalities of speech, sound, and music; (ii) flexible, on-demand thinking, allowing the model to do chain-of-thought-type reasoning before answering; (iii) multi-turn, multi-audio chat; (iv) long audio understanding and reasoning (including speech) up to 10 minutes; and (v) voice-to-voice interaction. To enable these capabilities, we propose several large-scale training datasets curated using novel strategies, including AudioSkills-XL, LongAudio-XL, AF-Think, and AF-Chat, and train AF3 with a novel five-stage curriculum-based training strategy. Trained on only open-source audio data, AF3 achieves new SOTA results on over 20+ (long) audio understanding and reasoning benchmarks, surpassing both open-weight and closed-source models trained on much larger datasets.