Updated

Nemotron 3.5 Bolt E4B 4GB

NVIDIA released Nemotron 3.5 Bolt E4B 4GB, an open small language model for conversational agents. NVIDIA describes multi-turn tool use, configurable reasoning, a 64K context window, and deployment on supported local NVIDIA hardware.

By No Code MBA · Facts checked

Provider
NVIDIA
Announcement date
Status at last check
Generally available
Model family
Nemotron 3.5
Preceding model
Nemotron 3 Nano 4B
Capabilities
Text

What changed

Nemotron 3.5 Bolt E4B 4GB is a text-in, text-out model built on Gemma 4 E4B (8B total, 4B effective parameters) with NVIDIA post-training for conversational agents. It adds multi-turn and asynchronous tool use, a concise speech-ready conversational mode, configurable reasoning with thinking caps, a 64K context window and support for 35 languages. NVIDIA reports about 40% higher average quality than Nemotron 3 Nano 4B and up to about 80% fewer output tokens than peers with reasoning on. NVIDIA lists about 106 to 133 tokens per second on an RTX 5070 at Q4_K_M.

Access and availability

Weights are on Hugging Face at nvidia/NVIDIA-Nemotron-3.5-Bolt-E4B-4GB under the OpenMDW-1.1 license (commercial use allowed), with GGUF, BF16 and DFlash drafter checkpoints. NVIDIA also lists distribution through Ollama, LM Studio, Unsloth, Jetson AI Lab and the In-Game Inferencing SDK.

Official access documentation

What you can build

Try it on a narrow local voice-assistant or tool-calling workflow, then measure response quality, latency, and memory use on your target hardware.

Limitations

Text only, with no vision input. NVIDIA's own tables show it trailing Qwen 3.5 4B on Tau2-Bench and RULER at 64K, and tight reasoning budgets can cut thoughts off mid-sentence. Asynchronous tool calling needs a specific acknowledgement pattern in the chat template.

Sources and release history