OpenAI, En Gelişmiş Dil Modeli GPT-4.5’i Tanıttı (Beklentilerinizi Karşılamayabilir…)

OpenAI has introduced its new language model GPT-4.5. The company claims it to be “the largest and most knowledgeable” model they have developed so far, yet it may not introduce revolutionary changes and could be less powerful than models like o1 or o3-mini.

GPT-4.5 will feature enhanced writing skills and a broader dataset, but according to OpenAI, it cannot be classified as a “groundbreaking” model. Leaked internal documents suggested that GPT-4.5 has achieved a tenfold increase in computational efficiency, though it may lack the expected new features.

Error Rate Lower than GPT-4o

It has been reported that the training of GPT-4.5 utilized OpenAI’s o1 model, codenamed Strawberry.

The company states that this new model is trained using techniques like supervised fine-tuning (SFT) and reinforcement learning from human feedback (RLHF). Its error rate is significantly lower than that of GPT-4o, although it will be slightly higher than the o1 model.

GPT-4.5 Will Better Understand Your Emotions

GPT-4.5 will be able to better perceive the underlying meaning of users’ writings in conversations or deduce intent from voice tone during spoken interactions.

Subscribers to OpenAI’s monthly $200 plan, ChatGPT Pro, will gain access to GPT-4.5 in ChatGPT starting this Thursday as part of a research preview. Developers under OpenAI’s paid subscription API options will also be able to use GPT-4.5 starting today.

Replay of GPT-4.5 Introduction Publication:

Recent leaks suggested that GPT-4.5 would similarly only be available to ChatGPT Pro users. OpenAI is expected to make a further announcement soon regarding when the model will be available to all users.