Free local AI from NVIDIA
Nemotron Hub: Free AI That Lives on Your Machine — Talks, and Now Listens
NVIDIA Nemotron 3.5 is the open-weight family that runs locally instead of billing per word. Bolt E4B (Oct 23, 2026) is a free 4GB conversational model that talks on almost any GPU. Transcribe (Oct 28, 2026) is a free speech model that turns recordings into speaker-labeled text in 40 languages, on your own NVIDIA GPU.
Bolt E4BLaunched Oct 23, 2026 — free, 4GB, Ollama in 15 minutes
TranscribeLaunched Oct 28, 2026 — free, 40 languages, up to 8 speakers
LicenseOpenMDW-1.1 — free to download, commercial use allowed
01
October launches
2Two Nemotron 3.5 models in one week: one talks, one listens.
№ 01nemotron transcribe guide
NVIDIA Nemotron 3.5 Transcribe Guide: Free AI That Turns Recordings Into Speaker-Labeled Text in 40 Languages
Free open-weight speech model: transcription, speaker labels, translation, audio Q&A on your own NVIDIA GPU. Setup path, limits, and the meeting-notes gig play.
New№ 02nemotron bolt e4b guide
NVIDIA Nemotron Bolt E4B Guide: A Free 4GB AI That Talks on Almost Any GPU
Free open-weight conversational model for local voice assistants, game characters, and robots. Ollama in 15 minutes, plus the $99/mo local voice-agent play.
New