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

Dinesh Manocha (University of Maryland, College Park) · Sreyan Ghosh (Nvidia) · Arushi Goel (NVIDIA Research) · Jaehyeon Kim (LG AI Research) · Sonal Kumar (Google / University of Maryland, College Park) · Zhifeng Kong (NVIDIA) · Sang-gil Lee (NVIDIA) · Chao-Han Yang (NVIDIA Research) · Ramani Duraiswami (University of Maryland) · Rafael Valle (NVIDIA) · Bryan Catanzaro (NVIDIA)
af-whisperaudio flamingo 3audio understanding benchmarksaudioskills-xlchain-of-thought reasoningclosed-source modelscurriculum-based training strategyjoint representation learninglarge audio-language modellarge-scale training datasetslong audio understandingmulti-turn audio chatopen-source audio datareasoning benchmarkssota resultsvoice-to-voice interaction

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.