News summaryTom's HardwareNov 14, 2026, 04:17 PM
Lawyer upgrades eight-V100 private AI server to run 80% of DeepSeek V4.1 Flash at 4-bit NVFP4 precision — adds 1TB of RAM, an Nvidia RTX 3090, and fixes crashes caused by a bad PSU
Nick Mermiges, founder of PLI Labs, upgraded the server to 1TB of RAM, added an RTX 3090, and moved about 80% of its workload from 2-bit to 4-bit NVFP4 precision. [[fact-jxhieas45]][[fact-ojr3gpb2x]]The server runs DeepSeek V4.1 Flash with a legal harness for his practice. [[fact-fg0pugiww]]Code output speed fell from 160–215 to 150–180 tokens per second. [[fact-a30kxnlvc]]The same-next-token measure rose from 92.2% to 95.1%. [[fact-se2nxg582]]