Artificial Intelligence (AI) has been a hot topic in recent years, with advancements in technology leading to smaller, better, and cheaper models that are challenging the traditional behemoths in the field. A recent state of the AI industry report revealed that 2024 was a breakthrough year for sleek, compact models that are giving the larger models a run for their money.
The report, released by the Institute for Human Centered AI at Stanford University, highlighted the rapid improvement in AI technology, with no single firm pulling ahead of the competition. The Chatbot Arena Leaderboard, which ranks the performance of various bots based on user votes, showed that the top-ranked model in early 2024 scored only 12% higher than the tenth-ranked model, but that margin had narrowed to just 5% by early 2025.
One of the key findings of the report was the trend towards smaller AI models that can match the performance of larger models from just a few years ago. Thanks to advancements in algorithms, modern models are now able to achieve the same level of performance as models that were 100 times larger in the past. This shift towards smaller models was described as a “breakthrough” in the field of AI.
Bart Selman, a computer scientist at Cornell University, praised the rise of smaller, more affordable AI models, noting that even small teams with innovative ideas could shake up the industry dominated by large companies. The report also highlighted the increasing competitiveness in the AI space, with most notable models now being developed by industry rather than academia.
While the United States has historically been the top producer of notable AI models, China has been rapidly catching up, producing models that rival their US counterparts in performance. The report also noted a surge in the number and performance of “open weight” models, which allow users to freely view the parameters that the models learn during training.
Despite the impressive advancements in AI technology, challenges such as implicit bias and false information generation still persist. While smaller models offer benefits such as faster training and lower energy consumption, there is still work to be done to address these issues.
Overall, the report paints a picture of a rapidly evolving AI landscape, with smaller, better, and cheaper models challenging the dominance of larger models. As the field continues to grow and evolve, it will be interesting to see how these trends shape the future of artificial intelligence.
