NVIDIA Releases Nemotron-3.5-Bolt-E4B-4GB Small Language Model
NVIDIA has released NVIDIA-Nemotron-3.5-Bolt-E4B-4GB, a small language model post-trained from Gemma for low-memory, low-latency edge and local voice assistant applications. The model enhances conversational capabilities, multi-turn tool use, and instruction following for embedded interfaces. This release provides developers with an efficient tool optimized for low-memory and fast time-to-first-token operation on edge devices. It enables faster response times for local voice assistants and interactive agents while preserving hardware headroom. The model is ready for commercial or non-commercial use and supports GGUF formats for local execution. Users can also utilize a companion DFlash drafter for speculative decoding or switch to the BF16 version for full precision.
## BACKGROUND
Speculative decoding is an inference-time optimization technique that uses a smaller draft model to propose multiple candidate tokens, which a larger target model verifies simultaneously to reduce latency. GGUF is an optimized file format designed to run large language models efficiently on local user hardware.