A curious new project has surfaced online, and it’s already sparking conversation among among researchers and coders. An open-source tool promises to track the “stupidity level” of major AI models in real time. While the name is tongue-in-cheek, the tool itself is serious—it’s designed to measure performance drops and help developers understand when popular models are cutting corners.
The tool, hosted at aistupidlevel.info, claims to be the first of its kind to monitor large language models for signs of decline. It currently tracks systems like OpenAI’s GPT-5 family, Anthropic’s Claude Opus 4, and Google’s Gemini 2.5 Pro, with support for xAI’s Grok 4 on the way.
Its approach is straightforward but wide-ranging: more than 140 coding and debugging tests run continuously, scoring models on correctness, stability, recovery, efficiency, and other factors. Results are fed into a live dashboard that shows how “smart” or “stupid” a model looks at any given time.
Another interesting element is cost analysis. The tool doesn’t just look at API pricing, but how many attempts a model needs to get something right. A supposedly “cheaper” model may waste cycles, while a more expensive one could finish the job faster and end up costing less overall.
Everything is open-source, with the code and API available on GitHub for anyone to review or contribute to. Since going live earlier this year, the site says it’s drawn nearly a million visitors, showing just how eager developers are for transparency in an increasingly closed-off industry.
Whether it’s a gimmick or a genuine accountability tool, the Stupid Meter highlights a growing frustration with AI performance swings. For developers and enthusiasts, it could become a useful way to separate hype from reality.
Don’t miss a thing! Join our Telegram community for instant updates and grab our free daily newsletter for the best tech stories!
For more daily updates, please visit our News Section.
(Source)
Comments