{"resourceId":"nvidia-nemotron-35-bolt-e4b-4gb-local-agent-model-20261023","versions":[{"version":"external-c0efd6570095b4391b43fb89a91cf5f30d0a5984489521fff0455b6f2cee1b30","resource":{"id":"nvidia-nemotron-35-bolt-e4b-4gb-local-agent-model-20261023","title":"NVIDIA releases Nemotron 3.5 Bolt E4B 4GB, an open small model post-trained from Gemma 4 E4B for local conversational agents on RTX, Jetson, RTX Spark and DGX Spark","organization":"NVIDIA","sector":"Technology vendor (open models for local and edge AI)","geography":"Global","publishedAt":"2026-10-23","publicationDate":"2026-10-23","eventDate":"2026-10-23","sourceName":"NVIDIA Research (Nemotron)","sourceLabel":"Vendor model release page","sourceUrl":"https://research.nvidia.com/labs/nemotron/Nemotron-3.5-Bolt-E4B-4GB/","evidenceClass":"vendor-claim","outcomeClass":"emerging","topics":["developers-agents","infrastructure","data-security"],"finding":"NVIDIA released Nemotron 3.5 Bolt E4B 4GB, an open small language model built on Google's Gemma 4 E4B with Nemotron post-training, aimed at voice and device assistants, interactive characters and robotic or IoT interfaces running on GeForce RTX, Jetson, RTX Spark and DGX Spark. NVIDIA claims about 40% higher average quality than its prior model on conversational-agent benchmarks, up to about 80% fewer output tokens than peers, operation in as little as about 4 GB in target workloads, multi-turn tool use, configurable reasoning, 64k context and support for 35 languages. It cites memory, latency, cost and privacy constraints as reasons developers choose local models.","sledRelevance":"Limited and indirect: small local models could support on-device assistants for public-facing kiosks, accessibility tools or field devices where data cannot leave the device, and teaching on campus RTX or Jetson hardware. No public-sector deployment is described.","evidence":"Vendor release with vendor-selected benchmarks measured on an RTX 5090; no independent evaluation on the page. See the Atomic Agent test for a small independent comparison.","architectureImplications":"Supports local inference on consumer and edge GPUs, keeping data on device; 64k context limit.","governanceImplications":"Derived from another vendor's base model; licence terms and model provenance need review before public-sector use.","securityPrivacyImplications":"Local execution can keep prompts and data on device, but tool-calling agents still need permission controls.","caveats":"Vendor claims; benchmarks chosen by NVIDIA; no SLED deployment evidence.","streamIds":["nvidia"],"roles":{"sales":"Sales framing: this is an NVIDIA model release, relevant to public-sector conversations only where on-device assistants or edge devices are on the table. Interpretation: privacy-sensitive uses such as kiosks, accessibility aids or field tools may benefit from models that run locally. Do not cite NVIDIA's benchmark gains as public-sector results; none exist. A bounded offer would be a short evaluation of local models on an agency's own tasks and hardware, paired with independent testing.","engineering":"Engineering framing: Bolt E4B is a Gemma 4 E4B derivative tuned for multi-turn tool use with a 64k context, aimed at RTX, Jetson and DGX Spark hardware. Pre-sales engineers should check licence terms, chat-template compatibility with existing agent frameworks, and actual memory use, since the 4 GB figure applies to target workloads. Test tool-calling permissions and failure behaviour on representative tasks before recommending it for any public-facing device.","delivery":"Delivery framing: adopting a new small model is mostly an evaluation and integration task. Interpretation: delivery teams would run side-by-side tests against the base model, confirm framework compatibility, document licence obligations and set guardrails for tool use. Licence obligations and the model's Gemma base should be recorded in the agency's model inventory. Acceptance criteria below are proposals for a pilot, not results reported by NVIDIA."},"roleDetails":{"sales":{"customerProblem":"Agencies want assistants that run on local devices without sending data to the cloud.","stakeholders":["Agency CIO","Accessibility coordinator","Privacy officer","Field-operations lead"],"discoveryQuestions":["Which tasks must run without network access?","What device hardware is available?","What licence terms are acceptable?","How will tool actions be authorised?"],"valueHypothesis":"Interpretation: local small models could reduce data exposure for simple assistant tasks; unproven in public sector.","potentialEngagement":"Local-model evaluation on agency tasks and hardware.","unsupportedClaims":["NVIDIA benchmark gains apply to agency workloads","Fits in 4 GB for all uses"]},"engineering":{"fit":"Edge or desktop devices with NVIDIA GPUs needing a small conversational agent.","architectureIntegration":"Local inference via llama.cpp, Ollama or similar on RTX/Jetson/DGX Spark; tool calls to local APIs.","prerequisites":["Compatible GPU","Agent framework support for the chat template","Licence review"],"constraints":["64k context limit","Vendor-only benchmarks","Small-model accuracy limits"],"security":"Restrict tool permissions and log actions; local execution limits data egress but not misuse of tools.","proofOfValue":["Compare with Gemma 4 E4B on agency tasks","Measure memory and latency on target devices"]},"delivery":{"work":"Evaluate, integrate and pilot the model with guardrails.","dependencies":["Framework compatibility","Licence approval","Test datasets"],"ownership":"Agency innovation or IT team.","skillsAdoption":"Staff need skills in local model serving and evaluation.","governanceCheckpoints":["Licence review","Privacy assessment","Pilot evaluation"],"acceptanceCriteria":["Proposed: meets task accuracy threshold on agency test set","Proposed: runs within device memory limits"],"risks":["Overstated vendor claims","Template or framework incompatibility","Unsafe tool actions"]}},"retrievedAt":"2026-10-24T02:55:59Z","enrichedAt":"2026-10-24T03:06:36Z","enrichmentBasis":"retrieved source","accessibilityWorkforceImplications":"Speech-ready conversational mode and multilingual support could matter for accessible public services; untested in that setting.","procurementImplications":"Open weights reduce licence cost but add review of licence and support arrangements.","operatingModelImplications":"Teams adopting the model would own evaluation, guardrails and updates locally rather than relying on a hosted service.","sourceVerification":{"openedUrl":"https://research.nvidia.com/labs/nemotron/Nemotron-3.5-Bolt-E4B-4GB/","referenceExcerpt":"Bolt E4B builds on Gemma 4 E4B with Nemotron post-training.","promptVersion":"sled-research-v4.0","model":null,"basis":"agent-reported inspection"}}}]}