Meta Launches GAMUT to Test AI Accuracy

Meta has released a new benchmark called GAMUT designed to measure how factually complete AI responses are. The system tests models using over one thousand multimodal questions to see if they provide accurate and thorough answers. This move highlights the ongoing struggle for tech companies to make artificial intelligence more reliable for daily tasks.

The results show that there is still a long way to go for the industry. Even the top performers struggled to pass the test with high scores. For example, Gemini 3.1 Pro only managed a score of 58.7 percent in the initial evaluations. These low numbers suggest that users should remain cautious when relying on AI for detailed or complex facts.

For the crypto and tech sectors, these benchmarks matter because accurate information is essential for data analysis and automated trading tools. As AI becomes more integrated into financial software, finding a model that does not hallucinate or miss key details is a top priority. We will be watching to see if other AI developers adopt these standards or if they build their own testing methods to improve their rankings.