Tips and Tutorials

Last week, OpenAI released the highly anticipated GPT-4o, the most powerful model to date. It is now available as an API on Azure, following its launch at Microsoft Build on May 21.Since then, our team has been dedicated to conducting regression tests on GPT-4o, and we are impressed with the results. We have identified significant improvements in quality, speed, and competitive cost. This provides us with exceptional capability for our real-time messaging business, with a speed close to that of GPT-35-Turbo.One of the great advantages of GPT-4o is its ability to adhere to the rules we set for it. This gives us much greater control over the model's responses, resulting in a significant reduction in so-called "hallucinations."Additionally, we are excited about a new feature we are experimenting with: GPT-4o allows us to understand images. This opens up a range of possibilities for our customers, making interaction with the model even richer and more diverse.The main benefits for our cust

Recently, two important events have stirred up the world of AI: Google I/O and OpenAI's Spring Update. Let's take a look at the key points from each of these events.Google had already launched Gemini 1.5 Pro with multimodal capabilities, but they had not yet been made available to everyone. Now, in OpenAI's Spring Update, they took the lead by releasing GPT-4o (the Omni version) with multimodal capabilities, surpassing Google in terms of model release.The launch of GPT-4o brought a slight increase in benchmark rankings, but it was enough to put OpenAI back in the lead as the most capable available model. Additionally, initial tests showed lower latency and lower cost compared to GPT-4 Turbo. The multimodal capacity of GPT-4o is truly impressive, allowing for more immersive interactions with the model and access to grounding with databases and API functions.At Google I/O, several product demonstrations were made showcasing AI with multimodal capabilities. As for the models, the highligh