GPT-4.5 has a greater emotional coefficient because it better understands the intentions and nuances of conversations

Openai has announced the launch of a previous view of its new GPT-4.5 language model, which has said that it is the “greatest and best for chatbot so far”, because it has the ability to recognize patterns, establish connections and generate knowledge, so that it better understands human needs and offers a greater “emotional coefficient.”

The company announced a few weeks ago that this model, known internally as Orion, would be its “last model of thought” and that it hoped that it was 100 times more powerful than GPT-4, “Using similar resourcesbut with an improved architecture and efficiency. ”

The new OpenAI model, GPT-4.5, is “one step forward in the extension of previous and posterior training” and has a broader knowledge base, as well as an improved capacity to follow the user’s intentions and a “greater emotional coefficient,” as the company commented in a statement.

Thanks to these skills, this model, which is offered at the moment as a preliminary version for research, has a deeper understanding of the world, which leads to reduce hallucinations and greater reliability in a wide range of issues. Therefore, it is from Usefulness for tasks such as improving writingprogramming and resolving practical problems.

Openai has also advanced that GPT-4.5 does not think before responding, “which makes its strengths particularly different from reasoning models, such as O1” and that with this model it has had the opportunity to develop new scalable techniques.

Thanks to this, it is possible to train larger and more powerful models with data derived from smaller models and can understand the nuances and natural conversations, which makes it the most large and best model for the chatbot so far. ”

To train the model, the company used artificial intelligence (AI) supercomputers of Microsoft Azure and new security techniques, which are combined with traditional supervised fine adjustment methods (SFT) and reinforcement learning from human feedback (RLHF), such as those used for GPT-4O.

GPT-4.5 It also has access to the most up-to-date information with the search, it asks File and image load And you can use Canvas to work in writing and code. At the moment, it does not admit multimodal functions, such as voice mode, video and shared screen in Chatgpt.

Openai has indicated that GPT-4.5 will reach all chatgpt pro-both users in the web version as in mobile apps and desktop- and will gradually expand for users of Plus y Team Next week. Finally, it will be implemented to subscribers of the version Enterprise y Edu, The following week.

Likewise, the brand has indicated that it is offering the previous view of the model in the application programming interface (API) of Chat completion, the API of attendees and the Lot API for developers at all levels of payment. From his experience, OpenAI has indicated that these professionals They can use it to plan and execute agentsincluding several steps coding flows and complex tasks automation.