AI Stupid Level is an independent, real-time benchmarking platform that scores large language models on coding, reasoning, tool-calling and speed, and detects performance drift over time. Below is the live model index.
AI providers update the model behind a stable API name without notice, so a model that scored well at launch may behave differently today. We benchmark continuously and apply CUSUM change-point detection to separate a genuine sustained decline from ordinary run-to-run noise. Read how AI drift detection and model degradation tracking works.
Every model runs the same three suites: a 7-axis code benchmark scored by executing the generated code, a deep reasoning benchmark covering multi-step problem solving and long-context retention, and a tool-calling benchmark measuring tool selection and argument accuracy. Scores are published with confidence intervals. See the full benchmarking methodology.
Put current models from OpenAI, Anthropic, Google, DeepSeek, Moonshot AI and Zhipu AI side by side on coding, reasoning, tool use, price and measured latency with the AI model comparison tool, or read the AI benchmarking FAQ.