Eseandre Mordi
⬤ Gemini 3 Preview (high) continues leading the Artificial Analysis Intelligence Index with a composite score of 73, tied with GPT-5.2 (high) for first place. The index measures performance across ten major benchmarks covering reasoning, coding, and general intelligence.
⬤ The latest rankings show Gemini 3 Preview (high) and GPT-5.2 (high) sharing the top spot with 73 points, while Claude Opus 4.5 and GPT-5.1 (high) follow closely at 70. Other top contenders like DeepSeek V3.2, Grok 4, and Claude Sonnet 4.5 cluster in the mid-60s. The tight score distribution shows just how competitive AI models have become, with small improvements making the difference between leaders.
⬤ The Intelligence Index combines results from ten independent tests including MMLU-Pro, GPQA Diamond, Humanity's Last Exam, LiveCodeBench, SciCode, AIME 2025, IFBench, AA-LCR, Terminal-Bench Hard, and T²-Bench Telecom. This approach smooths out the noise from single benchmarks and gives a clearer picture of overall capability. Gemini 3's consistent top performance points to well-rounded strength rather than excelling in just one area.
⬤ The results strengthen Google's standing as competition in frontier AI heats up. Ranking at the top of a widely-watched composite index matters when enterprises and developers are making decisions about which models to use. With leading systems now separated by just a few points, maintaining top scores on broad benchmarks shows how small gains can create meaningful advantages in the AI race.
Eseandre Mordi
