---
title: "Nemotron 3.5 Bolt E4B 4GB: An Open and... | SI Paper Desk"
url: https://thesimind.si/papers/2611.20771/
description: "NVIDIA's authors post-trained a small open model from Gemma-4-E4B-it for fast local assistants. They report the top average accuracy on 21 benchmarks and..."
---

Number 2 of 12 · November 18, 2026

# Nemotron 3.5 Bolt E4B 4GB: An Open and Efficient Small Language Model for Fast Conversational Agents

NVIDIA's authors post-trained a small open model from Gemma-4-E4B-it for fast local assistants. They report the top average accuracy on 21 benchmarks and faster decoding with a drafter model.

NVIDIA, :, Adi Margolin and 110 more · Submitted November 16, 2026 · arXiv 2611.20771v1 · cs.CL

[Read the paper on arXiv](https://arxiv.org/abs/2611.20771) · [The PDF](https://arxiv.org/pdf/2611.20771) · [The day's 12](https://thesimind.si/papers/day/2026-11-18/)

## What they found

The authors introduce Nemotron 3.5 Bolt E4B 4GB, a small open language model. It is post-trained from Gemma-4-E4B-it and tuned for speed and for intelligence per GB of memory. They say it improves multi-turn tool use, instruction following and input grounding (staying tied to the given input). On 21 benchmarks, they report the highest average accuracy and the fewest output tokens in both reasoning and non-reasoning modes, against Gemma-4-E4B-it and Qwen3.5-4B. The model ships with a DFlash drafter, a helper for speculative decoding (guessing ahead to speed up output). With it, the authors report 2.74× the per-user decode speed of inference without speculative decoding at concurrency 1. They also report 1.21× the generation throughput of Gemma-4-E4B-it with MTP at batch size 1.

## Why it matters

Small models that run on a local device can power voice assistants, interactive characters and robot or IoT interfaces. They need no round trip to a data center. The target hardware named is GeForce RTX, Jetson and RTX/DGX Spark. The abstract gives no details on the training method or the benchmarks, so the size of the gains is hard to judge from it alone. The comparisons are limited to two other small models.

## How it changes the race

This is mostly an incremental step in efficient small models. If the results hold, it makes capable local assistants cheaper and faster to run on modest hardware, and the open release lets more builders use it.

### The abstract, by the authors

We introduce Nemotron 3.5 Bolt E4B 4GB, an open and efficient small language model further post-trained from Gemma-4-E4B-it and optimized for speed and intelligence per GB of memory used for fast conversational agents. The model carries forward Gemma's strong conversational capabilities and further improves multi-turn tool use, instruction following, and input grounding for fast, local voice/device assistants, interactive personas, and embedded/Edge interfaces. The model combines natural conversation with multi-turn agentic capabilities to power voice/device assistants, interactive characters and robotic/IoT interfaces running on GeForce RTX, Jetson, and RTX/DGX Spark. Experiments on 21 benchmarks show that Nemotron 3.5 Bolt E4B 4GB achieves the highest average accuracy while generating the fewest output tokens in both reasoning and non-reasoning modes, compared with Gemma-4-E4B-it and Qwen3.5-4B. The model is released with a DFlash drafter. With DFlash, Bolt E4B achieves 2.74× the per-user decode speed of inference without speculative decoding at concurrency~1, and 1.21× the generation throughput of Gemma-4-E4B-it with MTP at batch size 1.

Quoted from arXiv:2611.20771. The research and its words belong to its authors.

Desk score **85** out of 100

Why it was picked

- Frontier lab: NVIDIA
- Agents (in the title)
- Reasoning
- Evals
- Language models (in the title)
- Studies frontier models (Qwen3.5)
- Cross-listed in 3 fields

Topics

- [Efficiency](https://thesimind.si/papers/?tag=efficiency)
- [Language models](https://thesimind.si/papers/?tag=language-models)
- [Agents](https://thesimind.si/papers/?tag=agents)
- [Benchmarks](https://thesimind.si/papers/?tag=benchmarks)
- [Training](https://thesimind.si/papers/?tag=training)
- [Speech](https://thesimind.si/papers/?tag=speech)

**arXiv ID:** 2611.20771v1
**Submitted:** November 16, 2026
**Fields:** cs.CL (Computation and Language), cs.AI (Artificial Intelligence), cs.LG (Machine Learning)
**Level:** Intermediate
**Comments:** Animated versions of Figure 3 are available in the blog: [https://research.nvidia.com/labs/nemotron/NVIDIA-Nemotron-3.5-Bolt-E4B-4GB/](https://research.nvidia.com/labs/nemotron/NVIDIA-Nemotron-3.5-Bolt-E4B-4GB/)

Fingerprint, SHA-256

`b3e25a7aa8a7e065ba8eadf2e633e08565f983c9447f3d7827493872734014bf`

On Solana: [Proof of Mind seal, 5WzZUk...UPMg](https://solscan.io/tx/5WzZUksJeCL3fYgfzcQ6DSdVxipu97PAmY5Y9M59gCMXbJtnyBVsqaoPSDHWrP4AiZAh7k4iVso5ts7WvxdNUPMg)

### The exact text fingerprinted

{"arxiv":"2611.20771","version":1,"title":"Nemotron 3.5 Bolt E4B 4GB: An Open and Efficient Small Language Model for Fast Conversational Agents","authors":["NVIDIA",":","Adi Margolin","Aditya Malte","Aishwarya Chhabra","Alex Hodson","Ameya Sunil Mahabaleshwarkar","Ambrish Dantrey","Amy Shen","Andrii Skliar","Annamalai Chockalingam","Anusha Pant","Ashwath Aithal","Ashwin Kumar","Besmira Nushi","Bhavesh Pawar","Bryan Catanzaro","Bryan Yu","Chantal Hwang","Chenhan Yu","Chris Alexiuk","Conner Currie","Cynthia Gao","David Mosallanezhad","Dina Yared","Ehsan Hosseini Asl","Eileen Long","Erick Galinkin","Felix Friedmann","Giedrius Burachas","Gorkem Batmaz","Grzegorz Chlebus","Guillermo Siman","Haifeng Qian","Hao Guo","Hayley Ross","Hrishith Thadicherla","Hui Wang","Huiying Li","Ike Nnoli","Irina Filatova","Iris Liu","Ji-Eun Lee","Jian Leong","Jian Zhang","Jiaqi Zeng","Joey Conway","Jonathan Cohen","Jose Vergara","Julien Veron Vialard","Kari Briski","Katherine Luna","Kedar Pandurang Potdar","Kevin Ning","Khalil Ben Khaled","Krzysztof Pawelec","Laura Dykes","Leili Tavabi","Lily Lee","Linda Liu","Liwei Ding","Maka Dong","Manoj Kilaru","Maria Bagdasarova","Matthew Lee","Meredith Price","Micah Schaffer","Michael Boone","Michael Fukuyama","Michael Hasper","Nancy Agarwal","Negar Habibi","Oleksii Kuchaiev","Oluwatobi Olabiyi","Pablo Ribalta","Payal Motwani","Pinky Xu","Prasoon Varshney","Pritam Biswas","Puhui Meng","Rameshwar Shivbhakta","Ran Zilberstein","Reka Kovacs","Richard Mazzarese","Rishabh Garg","Rita Fernandes Neves","Rochelle Woods","Ruby Ye","Ruvaid Virk","Ryan Hill","Saeid Motiian","Seph Mard","Shai Ben Nun","Shaona Ghosh","Sharad Yadav","Sudeep Sabnis","Suhas Hariharapura Sheshadri","Terry Kong","Thanh Do","Thinh Le","Udi Karpas","Viet Anh Trinh","Vinay Raman","Vinita Sharma","Warren Drones","Wenwen Gao","Whei Hsueh","Will Efstratis","Yoshi Suhara","Yunsheng Liu","Zhenming Wang","Zijia Chen","Zuhair Ahmed"],"day":"2026-11-18","rank":2,"score":85,"status":"written","summary":{"takeaway":"NVIDIA's authors post-trained a small open model from Gemma-4-E4B-it for fast local assistants. They report the top average accuracy on 21 benchmarks and faster decoding with a drafter model.","found":"The authors introduce Nemotron 3.5 Bolt E4B 4GB, a small open language model. It is post-trained from Gemma-4-E4B-it and tuned for speed and for intelligence per GB of memory. They say it improves multi-turn tool use, instruction following and input grounding (staying tied to the given input). On 21 benchmarks, they report the highest average accuracy and the fewest output tokens in both reasoning and non-reasoning modes, against Gemma-4-E4B-it and Qwen3.5-4B. The model ships with a DFlash drafter, a helper for speculative decoding (guessing ahead to speed up output). With it, the authors report 2.74× the per-user decode speed of inference without speculative decoding at concurrency 1. They also report 1.21× the generation throughput of Gemma-4-E4B-it with MTP at batch size 1.","matters":"Small models that run on a local device can power voice assistants, interactive characters and robot or IoT interfaces. They need no round trip to a data center. The target hardware named is GeForce RTX, Jetson and RTX/DGX Spark. The abstract gives no details on the training method or the benchmarks, so the size of the gains is hard to judge from it alone. The comparisons are limited to two other small models.","race":"This is mostly an incremental step in efficient small models. If the results hold, it makes capable local assistants cheaper and faster to run on modest hardware, and the open release lets more builders use it.","safety":"","level":"intermediate","tags":["efficiency","language-models","agents","benchmarks","training","speech"]}}

[Check it on Verify a thought](https://thesimind.si/proof-of-mind/)

## The authors

NVIDIA, :, Adi Margolin, Aditya Malte, Aishwarya Chhabra, Alex Hodson, Ameya Sunil Mahabaleshwarkar, Ambrish Dantrey, Amy Shen, Andrii Skliar, Annamalai Chockalingam, Anusha Pant, Ashwath Aithal, Ashwin Kumar, Besmira Nushi, Bhavesh Pawar, Bryan Catanzaro, Bryan Yu, Chantal Hwang, Chenhan Yu, Chris Alexiuk, Conner Currie, Cynthia Gao, David Mosallanezhad, Dina Yared, Ehsan Hosseini Asl, Eileen Long, Erick Galinkin, Felix Friedmann, Giedrius Burachas, Gorkem Batmaz, Grzegorz Chlebus, Guillermo Siman, Haifeng Qian, Hao Guo, Hayley Ross, Hrishith Thadicherla, Hui Wang, Huiying Li, Ike Nnoli, Irina Filatova, Iris Liu, Ji-Eun Lee, Jian Leong, Jian Zhang, Jiaqi Zeng, Joey Conway, Jonathan Cohen, Jose Vergara, Julien Veron Vialard, Kari Briski, Katherine Luna, Kedar Pandurang Potdar, Kevin Ning, Khalil Ben Khaled, Krzysztof Pawelec, Laura Dykes, Leili Tavabi, Lily Lee, Linda Liu, Liwei Ding, Maka Dong, Manoj Kilaru, Maria Bagdasarova, Matthew Lee, Meredith Price, Micah Schaffer, Michael Boone, Michael Fukuyama, Michael Hasper, Nancy Agarwal, Negar Habibi, Oleksii Kuchaiev, Oluwatobi Olabiyi, Pablo Ribalta, Payal Motwani, Pinky Xu, Prasoon Varshney, Pritam Biswas, Puhui Meng, Rameshwar Shivbhakta, Ran Zilberstein, Reka Kovacs, Richard Mazzarese, Rishabh Garg, Rita Fernandes Neves, Rochelle Woods, Ruby Ye, Ruvaid Virk, Ryan Hill, Saeid Motiian, Seph Mard, Shai Ben Nun, Shaona Ghosh, Sharad Yadav, Sudeep Sabnis, Suhas Hariharapura Sheshadri, Terry Kong, Thanh Do, Thinh Le, Udi Karpas, Viet Anh Trinh, Vinay Raman, Vinita Sharma, Warren Drones, Wenwen Gao, Whei Hsueh, Will Efstratis, Yoshi Suhara, Yunsheng Liu, Zhenming Wang, Zijia Chen, Zuhair Ahmed

Credit for this research belongs to its authors. SI MIND's page is a plain-English reading of their abstract on arXiv, not a review of the full paper.

Related on the desk

## More papers on these topics

[Nov 16, 2026 · Number 4LuthorSQL: A One-Pass Text-to-SQL Model as System 1 for a Slower LLM AgentA small local 7B model can answer nearly half of database questions in one fast pass and hand the hard ones to a big SI agent, cutting API cost by 33.3% at the same accuracy.EfficiencyAgentsLanguage models](https://thesimind.si/papers/2611.17780/) · [Nov 12, 2026 · Number 2AgentScript-Eval: Benchmarking LLMs and Agents for Code Generation in an Enterprise DSLOn a new benchmark for a private Salesforce agent language, SI models mostly fail zero-shot, and the setup around the model, like a compile-and-edit loop and clear guidance, mattered more than model size.BenchmarksCodeAgents](https://thesimind.si/papers/2611.13183/) · [Nov 11, 2026 · Number 10AMSchem: Automated Schematic Visualization of AMS Circuits via Verifiable Agentic GenerationAMSchem uses an SI agent with strict automatic checks to turn analog circuit netlists into readable schematics, reaching 85.0% validity versus 32.5% for direct generation on a 40-circuit subset.AgentsCodeEvals](https://thesimind.si/papers/2611.12333/)

[Number 1](https://thesimind.si/papers/2611.20983/) · [All 12 for Nov 18, 2026](https://thesimind.si/papers/day/2026-11-18/) · [Number 3](https://thesimind.si/papers/2611.20441/)
