Local LLM in OpenCode on 2x RTX 5090 | Autonomous Computer

Описание: One prompt produced a voxel Sakura Pagoda in a single HTML file, and the cost counter never moved.

The local LLM doing the work is Qwen3.8-27B-UD-Q8_K_XL, running under OpenCode on two RTX 5090. Nothing to meter, because the model is on the machine under the desk rather than in someone else's cloud. This is what 64GB of VRAM does between real jobs.

This is the Autonomous Computer 2, the smallest of the three. The same box takes two RTX PRO 6000 Blackwell instead, for 192GB, and ships with Ubuntu, CUDA, vLLM, SGLang and llama.cpp already on it. The Computer 4 takes four cards. The Computer 8 takes eight.

Own your compute. Own your intelligence.

Configure yours: https://www.autonomous.ai/computer-2
All three builds: https://www.autonomous.ai/datacenter
Build files: https://github.com/autonomous-ai/auto...

#OpenCode #Qwen #LocalLLM #LocalAI #RTX5090 #SelfHostedAI #AIWorkstation #AIServer #OnPremAI #LocalInference #AICodingAgent #AgenticCoding #VoxelArt #AutonomousComputer #PersonalAIComputer

Let's Connect:
Website: https://www.autonomous.ai
X: https://x.com/autonomous_labs
GitHub: https://github.com/autonomous-ai
Reddit:   / autonomouslounge  
Facebook:   / iamautonomous  
Instagram:   / autonomousdotai  
Tiktok:   / autonomous.ny