Benchmarks

2 articles

March 15, 2026

Google Gemini 3.1 Pro Claims the Reasoning Throne: Outperforming GPT-5.4 in Science and Logic

Google's latest AI model, Gemini 3.1 Pro, sets new world records in PhD-level science reasoning and logical pattern solving, challenging the dominance of OpenAI's GPT-5.4.

March 15, 2026

GPT-5.4 Released: OpenAI Breaks the Computer-Use Barrier with Human-Surpassing Performance

OpenAI's latest model, GPT-5.4, achieves a 75% success rate on OSWorld benchmarks, officially making it more proficient at computer desktop navigation than the average human expert.