2025 is seeing an acceleration in the race to achieve artificial superintelligence, fuelled by breakthroughs in large language models (LLMs). Tech giants and startups alike are launching increasingly advanced models, each striving to create AI systems that surpass human cognitive abilities.
Key advancements include OpenAI's GPT-5, DeepSeek's open-source DeepSeek-R1, and Google's Gemini 2.5 Pro. GPT-5 is touted as a major upgrade with impressive coding, maths, and science capabilities. DeepSeek-R1 uses reinforcement learning to achieve performance on par with OpenAI while reducing costs. Gemini 2.5 Pro excels in reasoning and coding. These models showcase improved reasoning, longer-term memory, and multimodality, processing text, images, and audio.
While some experts believe superintelligence is within reach, others argue that current systems are far from exceeding human performance across all cognitive domains. The pursuit of AGI and superintelligence is driving significant investment and innovation, with potential implications for various industries.
