Model WarsJuly 8, 2026via MarkTechPost
NVIDIA Releases Audex (Nemotron-Labs-Audex-30B-A3B): A Unified Audio-Text LLM That Preserves the Text Intelligence of Its Backbone
Why it matters
NVIDIA's Audex represents a significant capability expansion in multimodal LLMs—consolidating five audio/speech tasks into a single backbone while maintaining text performance. For enterprises, this signals a path toward unified inference endpoints that reduce deployment complexity and inference overhead.
Key signals
- Model: Nemotron-Labs-Audex-30B-A3B
- Architecture: MoE (Mixture of Experts)
- Capabilities: audio understanding, speech recognition, translation, TTS, audio generation
- Backbone: Nemotron-Cascade-2
- Text performance: marginal regression (claimed minimal loss)
- Publication date: July 8, 2026
The hook
NVIDIA just unified audio, speech, and text in one 30B model. Here's why the MoE architecture matters for inference costs.
NVIDIA's Nemotron-Labs-Audex-30B-A3B unifies audio understanding, speech recognition, translation, TTS, and audio generation in one MoE model. It keeps the text intelligence of its Nemotron-Cascade-2 backbone with marginal regression.
The post NVIDIA Releases Audex (Nemotron-Labs-Audex-30B-A3B): A Unified Audio-Text LLM That Preserves the Text Intelligence of Its Backbone appeared first on MarkTechPost.