Observed arrival · 2026-10-07
The Salaryman’s Local-AI Test Bench
A Japanese site measures local AI models’ speed and memory use, and links to a checker for whether they will run on your PC.
○Open
⊠Login
$Paid
†Ads
✦Pretty
●Pro
◎Niche
◉Human
⚑Risk
ƒJS
- For
- People comparing local AI models with their PC hardware
- Worth noticing
- The RTX 4080 table lists GPU and CPU speeds separately, including 216.9 versus 61.7 tokens per second for gemma3:1b.
Field notes
The benchmark table pairs each model with memory use, GPU and CPU throughput, a short speed description, and a measurement date. The featured results are labeled as RTX 4080 tests, while the PC-checker link points toward a separate hardware-matching function whose process is not visible here. The site describes its updates as fully automated by AI; its homepage also lists articles on topics including VRAM, Ollama, and local-model behavior.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
○OpenPublic substance visible
◎NicheUnusually specific use
One card from the complete issue