Skip to content
AI Model Radar

Nemotron Labs Audex 2B

nvidia/Nemotron-Labs-Audex-2B
Open Source · chat · open-weights
Open Alert me on changes
Context
128K
Max output
—
Weights
Open
API $/1M
—
Modalities
text
Released
06 Jul 2026
License: other · nvidia/Nemotron-Labs-Audex-2B
Download image Share on X Share on LinkedIn
AI summary
● machine-written

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

Nemotron-Labs-Audex-2B is a compact, dense audio-text language model that handles both audio and text tasks while preserving text-only reasoning capabilities. The model supports 128K-token context length and is trained using multi-stage supervised fine-tuning. It operates in both thinking and instruct modes for tasks including speech recognition, text-to-speech, audio generation, and text reasoning.

What's new
  • Smaller alternative to Audex-30B-A3B with reduced inference requirements
  • Supports 128K-token context window with ChatML template
  • Operates in both thinking and instruct (non-thinking) modes
  • Trained using multi-stage SFT recipe without cascaded RL
  • Handles audio understanding, speech recognition/translation, TTS, and audio generation
Best for
Speech recognition and translation tasksText-to-speech and audio generationAudio understanding and question-answeringMulti-modal reasoning with audio and text inputs
Sources

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