Claude Fable 5.2 from Anthropic leads SimpleBench, a benchmark that measures common-sense reasoning through multiple-choice questions. The model scores 94.9%, ahead of Google's Gemini 4 Carbon at 92.6% and Gemini 4 Argon at 91.7%. The score is published on the benchmark's own website, which presents it as the highest result recorded so far.
SimpleBench was built to find where language models fail on questions that a person with high school knowledge answers without effort. The questions cover spatio-temporal reasoning, social intelligence and so-called trick questions. The site lists more than 200 questions, but that number may change and it is not clear how many are actually evaluated.
Methodology is central to the comparison. The benchmark estimates the human level on a sample of nine participants, with a reference score of 83.7%, and on that basis argues that everyday reasoning remains a challenge for frontier models. The human figure rests on a small sample, and the site notes that model scores come from configurations tuned for the test.
The result fits a broader trend. According to Scale Labs and the Center for AI Safety, the same model also leads the Remote Labor Index, a test of real freelance jobs, where it completes 41.3%, compared with 2.5% a year ago. On the same benchmark its predecessor Fable 5.1 stays below 18%. Anthropic presented Fable 5.2 in early November as a direct response to GPT-6 Astra, with prices 20% lower and faster replies.
Google's picture differs. Gemini 4 Carbon was announced at the end of October as the company's strongest model, with leading results in cybersecurity, but for now it is available only to a restricted group of cyber defenders and selected partners. Reportedly, wider access will arrive only once protective systems are ready, with no specific date.
The gap between the SimpleBench scores is just over two percentage points. The value of the test lies mainly in showing where models still fail on elementary questions, a ground on which the top performers are still close.

No comments yet. Be the first!