Skip to content
AI Model Radar

Nemotron Labs Audex 2B

nvidia/Nemotron-Labs-Audex-2B
Open Source · chat · open-weights
GA Alert me on changes
Context
128K
Max output
—
Pricing
Open weights
Modalities
text
Released
06 Jul 2026
License: other · nvidia/Nemotron-Labs-Audex-2B
AI summary
● machine-written

NVIDIA releases Nemotron-Labs-Audex-2B, unified audio-text LLM

Nemotron-Labs-Audex-2B is a 2B-parameter audio-text language model that handles speech and audio understanding, generation, and reasoning while maintaining text capabilities. The model supports 128K-token context length and operates in both thinking and instruct modes. Released under a noncommercial license, it is available as open weights via Hugging Face.

What's new
  • Supports 128K-token context length for long-context tasks
  • Unified audio-text architecture preserves text reasoning performance
  • Operates in both thinking mode (with reasoning tags) and instruct mode
  • Handles audio understanding, speech recognition, text-to-speech, and audio generation
  • Available for local deployment via vLLM, Transformers, and SGLang
Best for
Audio understanding and speech recognition tasksText-to-speech and audio generationReasoning and long-context question answeringMulti-modal agent systems with audio inputs
Sources

Source: https://huggingface.co/nvidia/Nemotron-Labs-Audex-2B