Science
OpenAI's GPT-6 topped LiveBench and broke 1520 arena score
The model also outperformed all competitors in LiveBench Mathematics by the end of September.
OpenAI released GPT-6 on September 30, 2026, positioning it as a leading coding model. The model contributed to an AI system surpassing the 1520 Overall Arena score. Its performance secured the #1 spot on LiveBench Overall
The 1520 Overall Arena threshold was reached by an AI model before the end of September.
The next full Gemini model was evaluated for coding ability. It performed on par with Opus 4.7 and GPT-5.5 in benchmark tests. The comparison placed Gemini alongside the top coding models released this year. The assessment was completed by the end of September.
On LiveBench's Mathematics track, OpenAI's model ranked first at the close of September. Competing providers such as Anthropic, DeepSeek and Microsoft placed lower. The ranking reflected performance across a suite of mathematical problems.
These results collectively highlighted OpenAI's dominance across multiple benchmark categories. The simultaneous release of GPT-6 and the Gemini coding parity reinforced the company's competitive edge.
Analysis
Breaking the 1520 Overall Arena score and attaining the top LiveBench Overall ranking signal a notable leap in AI capability, likely encouraging broader deployment of OpenAI's models in commercial and research settings.
Sustaining this advantage will depend on continued innovation and scaling; uncertainties remain around hardware constraints, regulatory responses, and whether other organizations can close the gap in upcoming evaluations.
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.