Science
Anthropic topped every major AI benchmark in September
Anthropic led all major AI benchmarks in September while OpenAI placed Second
Anthropic finished September as the top performer on every major AI benchmark. The company led the rankings in general language modeling, Agent simulations, code generation, mathematical reasoning and Style-Control tasks. OpenAI ranked Second Overall, trailing Anthropic across the same suite of tests. Anthropic did not announce a solution to the Millennium Prize problem by the end of the month.
The benchmarks measured performance in five core areas. In general language tasks, Anthropic's model achieved the highest scores among all competitors. In Agent-based evaluations, its system again outperformed the field, followed by OpenAI's offering.
OpenAI remained the Second-best AI Agent Lab performer in September. The Agent Lab scores were part of the same benchmark suite evaluated at the end of the month.
Anthropic did not announce a Millennium Prize solution by September 30, 2026. The company confirmed the absence of a breakthrough on that date. The Prize therefore remained unclaimed at the close of the month.
Analysis
Anthropic's sweep of benchmarks indicated a broad technical advantage over rivals. Leading in general language, Agent, coding, math and Style-Control tasks suggested the company's models were versatile across diverse AI applications.
OpenAI's Second-place finish showed it retained strong capabilities but lagged behind Anthropic in the measured domains. The runner-up position may affect its standing among enterprise customers seeking top-tier AI performance.
Without a Prize-winning result, Anthropic's reputation remained tied to benchmark performance rather than a landmark theoretical advance.
What this was built from
Each claim above traces to a market below, at the price it was trading when this edition went to press.