NVIDIA ships Nemotron-3.5-Bolt-E4B, a 4GB Gemma-4-derived SLM for low-latency on-device voice assistants
jacek2023 · reddit · 2026-10-24
NVIDIA released Nemotron-3.5-Bolt-E4B (4GB) on Hugging Face, post-trained from Gemma-4-E4B-it. It keeps Gemma's conversational strength while improving multi-turn tool use, instruction following, and input grounding, targeting low-memory, low-TTFT use cases like local voice/device assistants, interactive personas, and embedded edge interfaces.
A companion DFlash drafter for speculative decoding, a BF16 full-precision version, and community GGUF quants (ggml-org, unsloth) are available. Licensed for commercial and non-commercial use.
More from Infra
- VC Rob Leclerc defends Big Tech mega-raises: 5-10% dilution buys far more growth — robleclerc · 2026-10-24
- NVIDIA's Hugging Face org ships open Nemotron lineup: 30B/3B hybrid MoE, 1M context, edge-ready — mervenoyann · 2026-10-24
- Google's TPU outsourcing could hit $100B-$250B by 2028, but HBM is the bottleneck — Beth_Kindig · 2026-10-24
- Spotify details LLM podcast recommendations: SGLang multi-item scoring cuts per-candidate request fan-out — ying11231 · 2026-10-24
- NVIDIA's Nemotron 3.5 Bolt lands on Ollama: small VLM with 128K context — ollama · 2026-10-24
- Bittensor Ecosystem Weekly: $30M in GPU Hosting at 91% Utilization, 10x Cheaper Video Model — markjeffrey · 2026-10-24