{"id":13194,"date":"2026-11-17T11:02:14","date_gmt":"2026-11-17T02:02:14","guid":{"rendered":"https:\/\/localmodelwatch.tsuchitsuchi.com\/?p=13194"},"modified":"2026-11-17T11:11:24","modified_gmt":"2026-11-17T02:11:24","slug":"nvidia-nemotron-3-5-bolt-e4b-gguf-2","status":"publish","type":"post","link":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/11\/17\/nvidia-nemotron-3-5-bolt-e4b-gguf-2\/","title":{"rendered":"NVIDIA-Nemotron-3.5-Bolt-E4B-GGUF Text Generation Model: 12GB+ VRAM"},"content":{"rendered":"<p><!-- lmw:facts --><\/p>\n<h2>At a Glance<\/h2>\n<div class=\"lmw-table-scroll\" tabindex=\"0\" style=\"overflow-x:auto;-webkit-overflow-scrolling:touch;max-width:100%;\">\n<table style=\"width:max-content;min-width:100%;border-collapse:collapse;\">\n<thead>\n<tr>\n<th>Item<\/th>\n<th>Value<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td>Repository<\/td>\n<td><a href=\"https:\/\/huggingface.co\/ggml-org\/NVIDIA-Nemotron-3.5-Bolt-E4B-GGUF\">ggml-org\/NVIDIA-Nemotron-3.5-Bolt-E4B-GGUF<\/a><\/td>\n<\/tr>\n<tr>\n<td>Family guide<\/td>\n<td><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/model-google-gemma-4-e4b-en\/\">gemma-4-E4B guide (2 articles)<\/a><\/td>\n<\/tr>\n<tr>\n<td>Publisher guide<\/td>\n<td><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/publisher-google-en\/\">Google: models and licenses<\/a><\/td>\n<\/tr>\n<tr>\n<td>Published<\/td>\n<td>2026-10-24<\/td>\n<\/tr>\n<tr>\n<td>License<\/td>\n<td>apache-2.0<\/td>\n<\/tr>\n<tr>\n<td>Formats<\/td>\n<td><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/format-gguf-en\/\">GGUF<\/a><\/td>\n<\/tr>\n<tr>\n<td>Source type<\/td>\n<td>Primary source (the publisher itself)<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n<p><em>Values determined by this site&#8217;s code when the information was collected. Dates are JST.<\/em><\/p>\n<p><!-- \/lmw:facts --><\/p>\n<h2>Overview<\/h2>\n<p>The GGUF format of NVIDIA&#8217;s small language model (SLM) &#8220;NVIDIA-Nemotron-3.5-Bolt-E4B&#8221; has been released by ggml-org. This model is based on Google&#8217;s &#8220;Gemma-4-E4B&#8221; with custom post-training applied by NVIDIA.<\/p>\n<p>The development aims to combine fast response times (low TTFT) and low memory consumption for locally running voice assistants, on-device assistants, interactive personas, and embedded\/edge interfaces. It is designed so that agents can react quickly within the limited resources of edge devices while reserving compute resources for applications and other model pipelines.<\/p>\n<p>Sources: <a href=\"https:\/\/huggingface.co\/ggml-org\/NVIDIA-Nemotron-3.5-Bolt-E4B-GGUF\">ggml-org\/NVIDIA-Nemotron-3.5-Bolt-E4B-GGUF<\/a> Sources: <a href=\"https:\/\/huggingface.co\/nvidia\/NVIDIA-Nemotron-3.5-Bolt-E4B-4GB-BF16\">nvidia\/NVIDIA-Nemotron-3.5-Bolt-E4B-4GB-BF16<\/a><\/p>\n<h2>Specifications<\/h2>\n<ul>\n<li>Parameters: 8B (4B effective parameters)<\/li>\n<li>Architecture: Gemma-4-E4B<\/li>\n<li>Context Length: Up to 64K tokens input, up to 64K tokens output<\/li>\n<\/ul>\n<h2>Performance<\/h2>\n<p>Regarding the performance of this model, benchmark results measured by the publisher NVIDIA are introduced below. The table below compares it with open-weight models of a similar scale: &#8220;Gemma4 E4B-it&#8221;, &#8220;Qwen3.5 4B&#8221;, and &#8220;Nemotron-3 Nano-4B&#8221;. Note that the figures are based on the pre-quantization BF16 version.<\/p>\n<div class=\"lmw-table-scroll\" tabindex=\"0\" style=\"overflow-x:auto;-webkit-overflow-scrolling:touch;max-width:100%;\">\n<p class=\"lmw-table-hint\" style=\"margin:0 0 4px;font-size:0.85em;opacity:0.7;\">\u2192 Scroll horizontally to see all columns<\/p>\n<table style=\"width:max-content;min-width:100%;border-collapse:collapse;\">\n<thead>\n<tr>\n<th style=\"text-align: left;\">Benchmark<\/th>\n<th style=\"text-align: center;\">Nemotron-3.5 Bolt-E4B<\/th>\n<th style=\"text-align: center;\">Gemma4 E4B-it<\/th>\n<th style=\"text-align: center;\">Qwen3.5 4B<\/th>\n<th style=\"text-align: center;\">Nemotron-3 Nano-4B<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td style=\"text-align: left;\">MultiChallenge<\/td>\n<td style=\"text-align: center;\">53.10<\/td>\n<td style=\"text-align: center;\">42.08<\/td>\n<td style=\"text-align: center;\">51.76<\/td>\n<td style=\"text-align: center;\">35.30<\/td>\n<\/tr>\n<tr>\n<td style=\"text-align: left;\">RoleBench<\/td>\n<td style=\"text-align: center;\">67.05<\/td>\n<td style=\"text-align: center;\">49.95<\/td>\n<td style=\"text-align: center;\">53.66<\/td>\n<td style=\"text-align: center;\">12.02<\/td>\n<\/tr>\n<tr>\n<td style=\"text-align: left;\">RoleMRC<\/td>\n<td style=\"text-align: center;\">80.63<\/td>\n<td style=\"text-align: center;\">74.62<\/td>\n<td style=\"text-align: center;\">65.81<\/td>\n<td style=\"text-align: center;\">47.00<\/td>\n<\/tr>\n<tr>\n<td style=\"text-align: left;\">EQ-Bench v3<\/td>\n<td style=\"text-align: center;\">39.40<\/td>\n<td style=\"text-align: center;\">41.17<\/td>\n<td style=\"text-align: center;\">32.92<\/td>\n<td style=\"text-align: center;\">19.31<\/td>\n<\/tr>\n<tr>\n<td style=\"text-align: left;\">BFCL-v3<\/td>\n<td style=\"text-align: center;\">68.51<\/td>\n<td style=\"text-align: center;\">66.39<\/td>\n<td style=\"text-align: center;\">70.95<\/td>\n<td style=\"text-align: center;\">61.10<\/td>\n<\/tr>\n<tr>\n<td style=\"text-align: left;\">Tau2-Bench<\/td>\n<td style=\"text-align: center;\">55.10<\/td>\n<td style=\"text-align: center;\">38.50<\/td>\n<td style=\"text-align: center;\">75.06<\/td>\n<td style=\"text-align: center;\">29.20<\/td>\n<\/tr>\n<tr>\n<td style=\"text-align: left;\">IHEval<\/td>\n<td style=\"text-align: center;\">84.79<\/td>\n<td style=\"text-align: center;\">67.04<\/td>\n<td style=\"text-align: center;\">60.57<\/td>\n<td style=\"text-align: center;\">51.31<\/td>\n<\/tr>\n<tr>\n<td style=\"text-align: left;\">AgentIF<\/td>\n<td style=\"text-align: center;\">25.81<\/td>\n<td style=\"text-align: center;\">21.89<\/td>\n<td style=\"text-align: center;\">20.44<\/td>\n<td style=\"text-align: center;\">20.08<\/td>\n<\/tr>\n<tr>\n<td style=\"text-align: left;\">StructEval-T<\/td>\n<td style=\"text-align: center;\">80.40<\/td>\n<td style=\"text-align: center;\">79.27<\/td>\n<td style=\"text-align: center;\">76.01<\/td>\n<td style=\"text-align: center;\">63.50<\/td>\n<\/tr>\n<tr>\n<td style=\"text-align: left;\">IFBench<\/td>\n<td style=\"text-align: center;\">63.79<\/td>\n<td style=\"text-align: center;\">42.84<\/td>\n<td style=\"text-align: center;\">58.91<\/td>\n<td style=\"text-align: center;\">60.10<\/td>\n<\/tr>\n<tr>\n<td style=\"text-align: left;\">RULER (64K)<\/td>\n<td style=\"text-align: center;\">86.99<\/td>\n<td style=\"text-align: center;\">86.55<\/td>\n<td style=\"text-align: center;\">95.15<\/td>\n<td style=\"text-align: center;\">91.51<\/td>\n<\/tr>\n<tr>\n<td style=\"text-align: left;\">LongBench v2 (48K)<\/td>\n<td style=\"text-align: center;\">50.50<\/td>\n<td style=\"text-align: center;\">49.34<\/td>\n<td style=\"text-align: center;\">59.93<\/td>\n<td style=\"text-align: center;\">31.40<\/td>\n<\/tr>\n<tr>\n<td style=\"text-align: left;\">GaRAGe<\/td>\n<td style=\"text-align: center;\">42.61<\/td>\n<td style=\"text-align: center;\">34.73<\/td>\n<td style=\"text-align: center;\">24.66<\/td>\n<td style=\"text-align: center;\">25.18<\/td>\n<\/tr>\n<tr>\n<td style=\"text-align: left;\">RAGTruth<\/td>\n<td style=\"text-align: center;\">77.59<\/td>\n<td style=\"text-align: center;\">76.97<\/td>\n<td style=\"text-align: center;\">75.60<\/td>\n<td style=\"text-align: center;\">71.40<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n<p>According to NVIDIA&#8217;s measurement results, this model records scores significantly higher than the comparison models, particularly in the areas of role-playing (RoleBench) and instruction-following (IHEval, IFBench). It also marks the highest values in RAGTruth, which measures RAG (Retrieval-Augmented Generation) accuracy, and AgentIF, which measures agent capabilities, demonstrating high suitability as an interactive agent.<\/p>\n<p>On the other hand, in EQ-Bench v3 which measures emotional intelligence, it falls slightly short of the base Gemma4 E4B-it, while Qwen3.5 4B records higher figures in function calling (BFCL-v3), complex agent tasks (Tau2-Bench), and long-context reading comprehension (RULER, LongBench v2). Particularly regarding long-context processing capability and tool-use accuracy, it is necessary to consider comparisons with competing models like Qwen3.5 4B depending on the task.<\/p>\n<p>For training, multi-environment reinforcement learning using GRPO (Group Relative Policy Optimization) was adopted. Through this, optimization was performed across instruction-following, multi-step tool use, multi-turn conversation, and structured output environments. This model is recommended not only for direct inference use but also as a base for domain-specific adaptation, distillation, and further post-training.<\/p>\n<h2>Strengths and Use Cases<\/h2>\n<p>This model is designed specifically for use as an AI agent running in local environments or an on-device assistant. In particular, it excels at processing complex tasks in the background while maintaining natural interactions with the user. As part of the base model&#8217;s design philosophy, it is optimized so that agents can react quickly (low TTFT) within the limited resources of edge devices while securing compute resources for applications and other processing pipelines.<\/p>\n<p>One of its major features is the inclusion of a &#8220;Thinking Mode&#8221; that can be controlled dynamically via chat templates. By setting <code>enable_thinking=True<\/code>, the model outputs an internal reasoning process (reasoning_content) before generating the final answer. This makes it possible to derive more accurate answers for complex questions requiring logical consistency. Additionally, a concept called &#8220;thinking budget&#8221; is introduced into this reasoning process, allowing applications to finely adjust the balance between response speed and accuracy by limiting the token amount spent on reasoning. The documentation also provides an implementation example of a two-stage inference client that first generates only the reasoning process and then generates the response from that continuation as needed.<\/p>\n<p>Furthermore, it features a &#8220;Conversational Mode&#8221; intended for use in voice dialogues and smart devices. Enabling this mode causes the model to omit redundant text decorations and formatting, generating concise, easy-to-hear responses suitable for text-to-speech (TTS). This is an extremely useful feature for use in screenless interfaces such as smart speakers and in-vehicle assistants.<\/p>\n<p>Regarding supported languages, while focusing primarily on English and coding languages, it officially supports multiple languages including Japanese, Spanish, French, German, and Italian. This allows Japanese users to build high-performance multilingual assistants in local environments. It also supports multi-turn Tool Calling, making it suitable for implementing autonomous agent functions linked with external APIs, such as fetching weather forecasts or performing calculation processing.<\/p>\n<p>Recommended sampling settings are Temperature 1.0 and Top_P 0.95, regardless of whether thinking mode is enabled or disabled. These settings maximize the model&#8217;s expressive power and reasoning capabilities.<\/p>\n<p>Specific anticipated use cases include:<br \/>\n&#8211; Locally running voice assistants and smart home device control<br \/>\n&#8211; Interactive character (persona) dialogue systems<br \/>\n&#8211; Advanced user interfaces on embedded systems and edge devices<br \/>\n&#8211; Building autonomous agents accompanied by multi-turn tool use<br \/>\n&#8211; Domain-specific adaptation or teacher models to transfer knowledge to smaller models<\/p>\n<h2>How It Differs from Similar Models<\/h2>\n<p>As a model based on Gemma-4-E4B similar to this model, there is the previously introduced <a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/2026\/10\/05\/google-diarizationlm-gemma-4-e4b-v1\/\">&#8220;DiarizationLM-Gemma-4-E4B-v1&#8221; Audio Processing Large Language Model: Required VRAM 8GB+<\/a>.<\/p>\n<p>The biggest difference between the two lies in the direction of their tuning. DiarizationLM is a model specialized in audio processing developed by Google, whose main purpose is &#8220;post-processing&#8221; to receive automatic speech recognition (ASR) results and perform speaker diarization or text formatting. In contrast, this Nemotron-3.5-Bolt-E4B is a model aimed at &#8220;interactive agents&#8221; where NVIDIA used reinforcement learning (RL) to enhance general instruction-following, complex reasoning, and tool-use capabilities. Even though they adopt the same lightweight base model, there is a clear difference in purpose: one aims to improve accuracy in specific audio tasks, while the other aims to achieve advanced general intelligence on local devices.<\/p>\n<p><!-- lmw:hardware --><\/p>\n<h2>Hardware Requirements<\/h2>\n<p><strong>Estimated requirements (calculated by Local Model Watch)<\/strong> \u2014 8.0B parameters (taken from the base model nvidia\/NVIDIA-Nemotron-3.5-Bolt-E4B-4GB-BF16)<\/p>\n<div class=\"lmw-table-scroll\" tabindex=\"0\" style=\"overflow-x:auto;-webkit-overflow-scrolling:touch;max-width:100%;\">\n<table style=\"width:max-content;min-width:100%;border-collapse:collapse;\">\n<thead>\n<tr>\n<th>Your VRAM<\/th>\n<th>Quantization<\/th>\n<th>File size<\/th>\n<th>Est. memory needed<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td>12GB (RTX 4070 \/ 3060 12GB, etc.)<\/td>\n<td>Q8_0<\/td>\n<td>7.5GB<\/td>\n<td>9.0GB<\/td>\n<\/tr>\n<tr>\n<td>24GB (RTX 4090 \/ 3090, etc.)<\/td>\n<td>BF16<\/td>\n<td>14.0GB<\/td>\n<td>16.8GB<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n<p><em>Memory estimates add a 20% runtime overhead (KV cache, etc.) to the actual size of the distributed files. Actual usage varies with context length, batch size and inference engine. These figures are computed by this site from file sizes, not published by the model&#8217;s authors. Compare with other models in our <a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/vram-guide-en\/\">VRAM quick reference<\/a>. What the quantization names mean: <a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/glossary-quantization-en\/\">glossary<\/a>.<\/em><\/p>\n<p><!-- \/lmw:hardware --><\/p>\n<p><!-- lmw:runnability --><\/p>\n<h2>Can You Run It Locally?<\/h2>\n<p><strong>Runs in Ollama, LM Studio and llama.cpp as-is.<\/strong><\/p>\n<p>It is distributed in GGUF, so no conversion is needed.<\/p>\n<p><strong>License \u2014 <code>apache-2.0<\/code> (Commercial use allowed):<\/strong> Permits commercial use, modification and redistribution. Redistribution requires including the license and stating changes; includes a patent grant.<\/p>\n<p><strong>Compression:<\/strong> the Q8_0 build measures 8.06 bits per weight \u2014 about 50% the size of the original 16-bit weights, calculated by this site from the actual file sizes.<\/p>\n<p><em>Compiled by this site&#8217;s code from the published formats, converted builds we have found, and each engine&#8217;s own model registry. &#8220;Not found&#8221; means we have not seen such a build, not that none exists. License summaries are not legal advice \u2014 check the publisher&#8217;s original terms before relying on them.<\/em><\/p>\n<p><!-- \/lmw:runnability --><\/p>\n<p><!-- lmw:peers --><\/p>\n<h2>Recent Models in the Same Size Class<\/h2>\n<p><em>Models with <\/em><em>4\u201315B<\/em><em> parameters that Local Model Watch covered recently, listed by code from our article log for comparison. VRAM tiers are this site&#8217;s estimates; licenses are as stated on the model cards.<\/em><\/p>\n<div class=\"lmw-table-scroll\" tabindex=\"0\" style=\"overflow-x:auto;-webkit-overflow-scrolling:touch;max-width:100%;\">\n<p class=\"lmw-table-hint\" style=\"margin:0 0 4px;font-size:0.85em;opacity:0.7;\">\u2192 Scroll horizontally to see all columns<\/p>\n<table style=\"width:max-content;min-width:100%;border-collapse:collapse;\">\n<thead>\n<tr>\n<th>Model<\/th>\n<th>Parameters<\/th>\n<th>Smallest VRAM tier<\/th>\n<th>License<\/th>\n<th>Our article<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td>microsoft\/UserRepGen-4B<\/td>\n<td>4.5B<\/td>\n<td>4GB<\/td>\n<td>mit<\/td>\n<td><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/11\/17\/userrepgen-4b\/\">UserRepGen-4B Text Generation Model: 4GB+ VRAM, GGUF Builds<\/a> (2026-11-17)<\/td>\n<\/tr>\n<tr>\n<td>IFM\/Jais-3-7B-GGUF<\/td>\n<td>9.0B<\/td>\n<td>8GB<\/td>\n<td>apache-2.0<\/td>\n<td><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/11\/17\/jais-3-7b-gguf-local-inference\/\">Jais-3-7B-GGUF Arabic and English 7B Reasoning Model: 8GB+ VRAM<\/a> (2026-11-17)<\/td>\n<\/tr>\n<tr>\n<td>IFM\/Jais-3-3.7B-GGUF<\/td>\n<td>5.1B<\/td>\n<td>4GB<\/td>\n<td>apache-2.0<\/td>\n<td><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/11\/17\/ifm-jais-3-37b-gguf\/\">Jais-3-3.7B-GGUF 512K Context Language Model: 4GB+ VRAM<\/a> (2026-11-17)<\/td>\n<\/tr>\n<tr>\n<td>ibm-granite\/granite-switch-4.1-8b<\/td>\n<td>8.8B<\/td>\n<td>24GB<\/td>\n<td>apache-2.0<\/td>\n<td><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/11\/17\/ibm-granite-switch-4-1-2\/\">IBM Releases Granite Switch 4.1: Modular LLMs with aLoRA and Speech<\/a> (2026-11-17)<\/td>\n<\/tr>\n<tr>\n<td>ByteDance-Seed\/Cola2<\/td>\n<td>8.8B<\/td>\n<td>48GB<\/td>\n<td>\u2014<\/td>\n<td><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/11\/17\/bytedance-seed-releases-cola2-language-model\/\">ByteDance-Seed Releases Cola2 Language Model<\/a> (2026-11-17)<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n<p><!-- \/lmw:peers --><\/p>\n<h2>How to Get It<\/h2>\n<p>This model is distributed in GGUF format and is available from the Hugging Face <code>ggml-org\/NVIDIA-Nemotron-3.5-Bolt-E4B-GGUF<\/code> repository. The Apache-2.0 license applies, allowing use for both commercial and non-commercial purposes.<\/p>\n<p>As an inference engine, operation is expected on llama.cpp and various tools utilizing it. The Hugging Face repository introduces the following command to launch a llama.cpp-based server:<\/p>\n<pre><code class=\"language-bash\">llama serve -hf ggml-org\/NVIDIA-Nemotron-3.5-Bolt-E4B-GGUF\n<\/code><\/pre>\n<p>Note that this model is not a gated model, so you can directly download and start trying it out without making special application requests or agreeing to terms with a Hugging Face account. While the original BF16 version (nvidia\/NVIDIA-Nemotron-3.5-Bolt-E4B-4GB-BF16) is also published, the GGUF version released this time is suitable for easy trial runs in local environments or operation with suppressed memory consumption.<\/p>\n<p><!-- lmw:related --><\/p>\n<h2>Related Articles<\/h2>\n<ul>\n<li><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/10\/05\/diarizationlm-gemma-4-e4b-v1\/\">DiarizationLM-Gemma-4-E4B-v1 Vision-Language Model: 8GB+ VRAM<\/a><\/li>\n<li><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/11\/17\/qwen3-coder-30b-a3b-instruct-gguf-2\/\">Qwen3-Coder-30B-A3B-Instruct-GGUF Text Generation Model: 24GB+ VRAM<\/a><\/li>\n<li><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/11\/17\/ggml-org-flint-gguf\/\">Flint-GGUF Coding MoE Language Model: 32GB+ VRAM<\/a><\/li>\n<li><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/11\/17\/vllm-d3-gguf-models-released\/\">vllm-d3 GGUF Models Released: Decision Models for Local Run<\/a><\/li>\n<\/ul>\n<p><!-- \/lmw:related --><\/p>\n<p><!-- lmw:next-steps --><\/p>\n<h2>What to Read Next<\/h2>\n<ul>\n<li><strong>Find models by VRAM<\/strong> (This model runs from the 12GB tier) \u2192 <a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/vram-guide-12gb-en\/\">Other models that run on a 12GB GPU<\/a><\/li>\n<li><strong>Explore the same model family<\/strong> \u2192 <a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/model-google-gemma-4-e4b-en\/\">gemma-4-E4B family overview (2 articles, 1 converted builds)<\/a><\/li>\n<li><strong>Engines that run this model<\/strong> \u2192 <a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/engine-llama-cpp-en\/\">llama.cpp<\/a> \/ <a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/engine-ollama-en\/\">Ollama<\/a><\/li>\n<li><strong>What Q8_0 mean and where to get this model<\/strong> \u2192 <a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/format-gguf-en\/\">GGUF format guide and models<\/a> \/ <a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/glossary-quantization-en\/\">Quantization and model-format glossary<\/a><\/li>\n<li><strong>Learn about the publisher<\/strong> \u2192 <a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/publisher-google-en\/\">Google: models, licenses and articles<\/a><\/li>\n<\/ul>\n<p><!-- \/lmw:next-steps --><\/p>\n<h2>Sources<\/h2>\n<ul>\n<li><a href=\"https:\/\/huggingface.co\/ggml-org\/NVIDIA-Nemotron-3.5-Bolt-E4B-GGUF\">https:\/\/huggingface.co\/ggml-org\/NVIDIA-Nemotron-3.5-Bolt-E4B-GGUF<\/a><\/li>\n<li><a href=\"https:\/\/huggingface.co\/nvidia\/NVIDIA-Nemotron-3.5-Bolt-E4B-4GB-BF16\">https:\/\/huggingface.co\/nvidia\/NVIDIA-Nemotron-3.5-Bolt-E4B-4GB-BF16<\/a><\/li>\n<li><a href=\"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\/<\/a><\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"<p>Discover NVIDIA-Nemotron-3.5-Bolt-E4B in GGUF format by ggml-org. A fast, low-memory SLM based on Gemma-4-E4B for local AI agents and assistants.<\/p>\n","protected":false},"author":1,"featured_media":13193,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[310],"tags":[2982,161,163,1073,4073,1547,1555,4075],"class_list":["post-13194","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-new-models","tag-gemma-4-e4b-en","tag-ggml-en","tag-gguf-en","tag-llama-cpp-en","tag-nvidia-nemotron-3-5-bolt-e4b-en","tag-verified","tag--en"],"lang":"en","translations":{"en":13194,"ja":13192},"_links":{"self":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/13194","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/comments?post=13194"}],"version-history":[{"count":2,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/13194\/revisions"}],"predecessor-version":[{"id":17929,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/13194\/revisions\/17929"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media\/13193"}],"wp:attachment":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media?parent=13194"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/categories?post=13194"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/tags?post=13194"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}