{"version":1,"type":"story","url":"https://digestai.news/story/nvidia-releases-nemotron-3-5-bolt-e4b-for-local-voice-agents-and-robot","json":"https://digestai.news/story/nvidia-releases-nemotron-3-5-bolt-e4b-for-local-voice-agents-and-robot.json","markdown":"https://digestai.news/story/nvidia-releases-nemotron-3-5-bolt-e4b-for-local-voice-agents-and-robot.md","slug":"nvidia-releases-nemotron-3-5-bolt-e4b-for-local-voice-agents-and-robot","headline":"NVIDIA releases Nemotron 3.5 Bolt E4B for local voice agents and robots","summary":"NVIDIA released Nemotron 3.5 Bolt E4B 4GB, an open small language model designed for fast conversational agents running locally on GeForce RTX, Jetson, RTX Spark, and DGX Spark hardware. Post-trained exclusively from Google's Gemma 4 E4B using reinforcement learning rather than supervised fine-tuning, the model features 8B total parameters, 4B effective parameters, and a 64K token context window. In NVIDIA tests, a quantized Q4_K_M build consumed between 3.1 and 4.3 GB of VRAM.\n\nAccording to NVIDIA's internal benchmarks across 21 evaluations, Bolt outperformed Gemma 4 E4B, Qwen3.5-4B, and Nemotron 3 Nano 4B on 15 tests in reasoning mode. It reached 84.8 on IHEval compared to Gemma's 67.0, though it trailed Qwen3.5-4B on Tau2-Bench with 55.1 versus 75.1. Using speculative decoding with a DFlash drafter on an RTX 5070, NVIDIA reports generation speeds reaching 272.6 tokens per second.\n\nThe model is licensed under OpenMDW-1.1 and is distributed on Hugging Face, Ollama, LM Studio, Unsloth, and Jetson AI Lab, alongside NVIDIA NIM and the In-Game Inferencing SDK. Although named '4GB' to reflect its memory footprint goal, the downloadable Q4_K_M GGUF file measures 6.18 GB.","keyPoints":["NVIDIA launched Nemotron 3.5 Bolt E4B with 8B total and 4B effective parameters, fitting in roughly 4GB VRAM.","Trained from Google's Gemma 4 E4B using reinforcement learning without supervised fine-tuning, featuring a 64K context window.","NVIDIA claims it won 15 of 21 internal benchmarks against competing small models, reaching 272.6 tokens/s on an RTX 5070."],"whyItMatters":"Fitting conversational reasoning into roughly 4GB of VRAM allows developers to run speech, vision, and language agents simultaneously on consumer GPUs and edge devices.","category":{"slug":"models","name":"Generative AI & Models","url":"https://digestai.news/category/models"},"entities":{"companies":["NVIDIA","Google","Scale Labs","Hugging Face"],"models":["Nemotron 3.5 Bolt E4B","Gemma 4 E4B","Qwen3.5-4B","Nemotron 3 Nano 4B","Cosmos 3 Edge"],"people":["Asif Razzaq"]},"firstPublishedAt":"2026-10-24T03:56:29Z","updatedAt":"2026-10-24T03:56:29Z","sourceCount":2,"hasPrimarySource":false,"sources":[{"outlet":"MarkTechPost","title":"NVIDIA Releases Nemotron 3.5 Bolt E4B 4GB: An Open Small Language Model for Local Voice Agents, Game Characters and Robots","url":"https://marktechpost.com/2026/10/23/nvidia-nemotron-3-5-bolt-e4b-4gb","publishedAt":"2026-10-24T03:56:29Z","type":"press","primary":false,"lead":true},{"outlet":"note.com","title":"With the arrival of NVIDIA Nemotron 3.5 Bolt E4B, those who delegate small tasks to local AI will benefit","url":"https://note.com/unikoukokun/n/n0106026fc993?hl=en","publishedAt":"2026-10-23T20:25:00Z","type":"press","primary":false,"lead":false}],"sourceNotes":null,"discussions":[],"thread":{"title":"AI Model Showdown Coding to Voice","url":"https://digestai.news/thread/recursive-releases-flint-40b-coding-agent-model-on-hugging-face","storyCount":2},"cite":{"text":"Digest AI, \"NVIDIA releases Nemotron 3.5 Bolt E4B for local voice agents and robots\", 24 October 2026, https://digestai.news/story/nvidia-releases-nemotron-3-5-bolt-e4b-for-local-voice-agents-and-robot","publisher":"Digest AI","title":"NVIDIA releases Nemotron 3.5 Bolt E4B for local voice agents and robots","datePublished":"2026-10-24T03:56:29Z","url":"https://digestai.news/story/nvidia-releases-nemotron-3-5-bolt-e4b-for-local-voice-agents-and-robot"},"generatedBy":"Written by Digest AI's editorial model from the linked sources; the sources are the record.","license":"Headlines, digests and key points are written by Digest AI and may be quoted with a link to the story page. Linked articles belong to their publishers. Terms: https://digestai.news/terms#reuse"}