Local text AI: running Nemotron on a home PC
Nemotron-Cascade-2-30B-AWQ ran locally on a home setup with a GeForce RTX 3090, delivering 120–150 tokens per second, and up to 210+ when counting reasoning tokens. In tests, the model handled physics, biology and Python tasks with confidence, as well as web agents. The main catch turned out not to be a hardware limit but handling the reasoning blocks correctly through vLLM.
AI-processed from Habr AI; edited by Hamidun News
It turns out you can run the 30-billion-parameter Nemotron-Cascade-2 at home on a GeForce RTX 3090 and use it as a local AI for working with text, coding and data analysis at 120–150 tokens/s.
Which neural network runs locally on an RTX 3090?
Nemotron-Cascade-2 is a local model that produces 120–150 tokens per second on a GeForce RTX 3090 and handles coding, physics, biology and web agents.
Which neural network can you run on an RTX 3090 graphics card at home?
Nemotron-Cascade-2 (30 billion parameters) runs locally on an RTX 3090 at 120–150 tokens per second with a stable context, providing reasoning good enough for coding and analytical tasks.
Why use a local neural network on a GPU instead of cloud services?
Running locally gives you fast responses (150 tokens/s on an RTX 3090), a long context and full control over the model for everyday work, without cloud latency.
Can you run Nemotron on a home PC?
Yes, the 30-billion-parameter Nemotron-Cascade-2 can be used on a home PC with a GeForce RTX 3090 at 120–150 tokens/s.
What tasks is the local Nemotron model suited to?
Nemotron handles coding, physics and biology problems, and runs web agents without the cloud.
- Nov 1, 2026Local models lose to Opus and Codex: 0.21 on a MacBook versus 1.00 for Codex
- Nov 1, 2026Kimi scores 96.2% on a 1C database analytics test and emerges as a fallback if Claude access is blocked
- Nov 1, 2026OpenAI Rewrites GPT Behavior Rules: Five Values and a Chain of Command
- Oct 31, 2026Author puts his script before an AI jury: the verdict cost 850 thousand tokens
Key Questions Answered
Which neural network runs locally on an RTX 3090?
Nemotron-Cascade-2 is a local model that produces 120–150 tokens per second on a GeForce RTX 3090 and handles coding, physics, biology and web agents.
Which neural network can you run on an RTX 3090 graphics card at home?
Nemotron-Cascade-2 (30 billion parameters) runs locally on an RTX 3090 at 120–150 tokens per second with a stable context, providing reasoning good enough for coding and analytical tasks.
Why use a local neural network on a GPU instead of cloud services?
Running locally gives you fast responses (150 tokens/s on an RTX 3090), a long context and full control over the model for everyday work, without cloud latency.
Can you run Nemotron on a home PC?
Yes, the 30-billion-parameter Nemotron-Cascade-2 can be used on a home PC with a GeForce RTX 3090 at 120–150 tokens/s.
Test Frontier LLMs & Autonomous Agents in One Workspace
Run Claude Opus 5.5, GPT-6, Gemini 3.1, and DeepSeek side-by-side in Nomad AI (nomadnet.ai) — unified model routing, deep research, and multimodal workflows under one subscription.
The AI world, distilled — once a week
Seven stories that actually mattered, hand-picked. No noise, no reposts, no press releases.
Done! Check your inbox for a confirmation.