NVIDIA has released Nemotron 3.5 Bolt E4B 4GB, an open small language model optimized for fast conversational agents. This model is designed to power voice assistants, game characters, and robot interfaces operating locally on NVIDIA's RTX, Jetson, and Spark hardware.
Key points
- NVIDIA released Nemotron 3.5 Bolt E4B 4GB, an open small language model.
- The model targets local voice agents, game characters, and robots.
- It is post-trained from Google's Gemma 4 E4B using reinforcement learning.
- Nemotron 3.5 Bolt E4B requires approximately 4GB of VRAM.
- It achieved top scores on 15 of 21 benchmarks against competitors.
The model is post-trained from Google's Gemma 4 E4B using reinforcement learning. It requires approximately 4GB of VRAM, leaving additional space for speech and vision models. Nemotron 3.5 Bolt E4B features 8 billion total parameters with an effective 4 billion and a 64K context window.
In NVIDIA's tests, the Q4_K_M build utilized between 3.1 to 4.3 GB of VRAM on RTX hardware. Performance benchmarks show Nemotron 3.5 Bolt E4B achieving the top score on 15 out of 21 benchmarks when compared against Gemma 4 E4B, Qwen3.5-4B, and Nemotron 3 Nano 4B.
By the numbers
- VRAM requirement
- 4GB
- total parameters
- 8B
- context window
- 64K
Nemotron 3.5 Bolt E4B enables multi-turn tool calling and supports asynchronous tool calls. It includes a conversational_mode flag for short, speech-ready responses and configurable reasoning with a runtime thinking budget to manage latency. The model also supports a 64K context for history, tool definitions, and application state.
NVIDIA opted to skip supervised fine-tuning for Bolt E4B, instead employing RLVR and RLHF through NeMo-RL and NeMo-Gym. This approach aimed to improve tool calling while preserving Gemma's conversational quality and verbosity.
Sources and references
Original report
NVIDIA Releases Nemotron 3.5 Bolt E4B 4GB: An Open Small Language Model for Local Voice Agents, Game Characters and Robots
MarkTechPost
Read on MarkTechPost (opens in a new tab)Linked in the original report
- Linkresearch.nvidia.com/labs/nemotron/NVIDIA-Nemotron-3.5-Bolt-… (opens in a new tab)
- Modelhuggingface.co/google/gemma-4-E4B-it (opens in a new tab)
- Modelhuggingface.co/nvidia/NVIDIA-Nemotron-3.5-Bolt-E4B-BF16 (opens in a new tab)
- Paperresearch.nvidia.com/labs/nemotron/files/NVIDIA-Nemotron-3.5… (opens in a new tab)
- Linkblogs.nvidia.com/blog/local-ai-rtx-spark-laptops-softwar… (opens in a new tab)
- Codegithub.com/NVIDIA-NeMo/RL (opens in a new tab)
Ref 00BBW8DF
Cite this
APA
Interestana AI Editorial. (2026, October 24). NVIDIA Releases Nemotron 3.5 Bolt E4B Open Language Model. Interestana. https://interestana.com/articles/nvidia-releases-nemotron-35-bolt-e4b-4gb-an-open-small-language-model-for-local-voice-agents-game-characters-and-robots-00bbw8df
MLA
“NVIDIA Releases Nemotron 3.5 Bolt E4B Open Language Model.” Interestana, 24 Oct. 2026, interestana.com/articles/nvidia-releases-nemotron-35-bolt-e4b-4gb-an-open-small-language-model-for-local-voice-agents-game-characters-and-robots-00bbw8df.
Chicago
Interestana AI Editorial. “NVIDIA Releases Nemotron 3.5 Bolt E4B Open Language Model.” Interestana, October 24, 2026. https://interestana.com/articles/nvidia-releases-nemotron-35-bolt-e4b-4gb-an-open-small-language-model-for-local-voice-agents-game-characters-and-robots-00bbw8df.
BibTeX
@misc{interestana_00bbw8df,
author = {{Interestana AI Editorial}},
title = {{NVIDIA Releases Nemotron 3.5 Bolt E4B Open Language Model}},
howpublished = {Interestana},
year = {2026},
month = oct,
url = {https://interestana.com/articles/nvidia-releases-nemotron-35-bolt-e4b-4gb-an-open-small-language-model-for-local-voice-agents-game-characters-and-robots-00bbw8df},
note = {Summary of reporting by MarkTechPost, https://www.marktechpost.com/2026/10/23/nvidia-nemotron-3-5-bolt-e4b-4gb/}
}Credit the source
Original reporting: MarkTechPost, “NVIDIA Releases Nemotron 3.5 Bolt E4B 4GB: An Open Small Language Model for Local Voice Agents, Game Characters and Robots”, https://www.marktechpost.com/2026/10/23/nvidia-nemotron-3-5-bolt-e4b-4gb/(opens in a new tab)
Follow this story
Get Artificial Intelligence in your inbox — new stories in 40 words or fewer, once a week.
Subscribe