Oriel — AI & technology worth noticing, every day.

unsloth/NVIDIA-Nemotron-3.5-Bolt-E4B-GGUF

AI model · https://huggingface.co/unsloth/NVIDIA-Nemotron-3.5-Bolt-E4B-GGUF

First seen by Oriel on .

AI summary

The page documents NVIDIA-Nemotron-3.5-Bolt-E4B-4GB, a Q4_K_M model post-trained from Gemma-4-E4B-it for low-memory local assistants, with emphasis on multi-turn tool use and instruction following.

  • The model has 8B total parameters, 4B effective parameters, and supports contexts up to 64K tokens.
  • It was trained with multi-environment GRPO reinforcement learning across instruction following, multi-step tool use, multi-turn conversations, and structured output.
  • Deployment examples use llama.cpp with an optional DFlash drafter for speculative decoding; chat-template controls toggle reasoning and conversational modes.
  • Benchmark results vary by task: the quantized model scores 74.13 on ToolSandbox and 85.76 on RULER (64K), while Qwen3.54B Q4_K_M scores 74.51 and 95.37, respectively.
  • The model is intended for local voice and device assistants, interactive personas, and embedded interfaces, and is governed by OpenMDW License Agreement version 1.1.

#language models #llama.cpp #speculative decoding #reinforcement learning

Generated by gpt-5.6-sol. Check the original for details.

Metric trends

Last 30 days · Up to 120 recorded samples per source and metric, with reposts kept separate · First and latest retained

Likes

Total likes reported by the source.

  • Hugging Face Models: 153, latest observation ; +43 over observed period since

Downloads (30d)

Rolling 30-day downloads. A change in this value is not new downloads during the observed period.

  • Hugging Face Models: 162,308, latest observation ; +122,387 over observed period since

Aliases

SourceExternal IDURL
Hugging Face Modelsunsloth/NVIDIA-Nemotron-3.5-Bolt-E4B-GGUFhttps://huggingface.co/unsloth/NVIDIA-Nemotron-3.5-Bolt-E4B-GGUF Read discussion