OpenAI is competing with the wrong LLM! Gemini is currently preferred 10x less by customers than Anthropic Sonnet 3.5 still remains one of the most popular LLMs while GPT-4o seems to waning The o1 line is kinda cool but it is expensive and slow because of which it's not…
Ok the new GPT-4o is WAY better
OpenAI met à jour GPT-4o et reprend sa couronne de meilleur modèle d'IA https://t.co/4ehT8WhB8F
OpenAI has released an updated version of its GPT-4o model, which shows mixed performance results compared to its previous iteration from August 2024. Key metrics indicate a decrease in the Artificial Analysis Quality Index from 77 to 71, and a drop in GPQA Diamond scores from 51% to 39%. Additionally, the model's performance in mathematical tasks has declined, with scores falling from 78% to 69%. However, the new version has demonstrated improved speed, increasing output from approximately 80 tokens per second to 180 tokens per second, and enhancements in creative writing capabilities. The competition in the AI landscape remains intense, particularly with Google’s Gemini model, as analysts note that forthcoming large models from OpenAI, Google, and Anthropic have not met expected performance gains despite advancements in training data and computing power. Some users have expressed dissatisfaction with the updates, suggesting that the constant adjustments could be detrimental to the model's overall quality.
