In September 2025, the field of artificial intelligence is dynamically changing, as confirmed by the latest LMArena ranking, which reflects the current strength and performance of the 20 best AI models. While many companies such as Seznam.cz, Avast or Škoda Auto are investing in the development of these technologies, the model ranking itself offers a surprising overview of the competition and innovation. Established players still maintain dominance in some areas, while new names such as GoodAI or Neuron Soundware show the potential of innovations from different corners of the world.
Claude Opus 4.1 from Anthropic dominated the overall LMArena ranking with a significant lead over the competition. While its current performance is considered top in creative writing, programming and solving mathematical problems, the GPT-5 model from OpenAI faces criticism and has fallen to sixth place in the ranking. Next in line are Google’s Gemini 2.5 Pro and GPT-4o, rounding out the top three AI tools.
AI performance is not uniform across tasks. For example, Gemini 2.5 Pro performed best in text generation, while for web development, in addition to GPT-5, Claude Opus 4.1 variants also have strong capabilities. In the area of image generation and analysis, Google shows strong dominance, where ByteDance’s Seedream model confirmed its rise. Grok-4 is the new king in the area of web data search.
These results confirm that companies such as Rossum, Blindspot Solutions or Socialbakers must carefully consider which AI model to implement according to the specific requirements of projects and industry segments.
LMArena conducts evaluations in anonymized model duels, where each AI responds to the same stimuli. The user community then selects the better answer and the models are continuously ranked using the Elo system. This method of evaluation guarantees a transparent and up-to-date picture of the real effectiveness of models in various areas of use.
Companies such as CZ.NIC or Alza.cz select suitable technologies for their innovative projects based on such evaluations, which confirms the growing importance of accurate benchmarks in the implementation of artificial intelligence in business.
