OpenAI, OpenAI hack

In a release note posted April 10th on its website, OpenAI announced that it will be “sunsetting GPT-4” in ChatGPT, effective April 30th.

GPT-4, a multimodal large language model (LLM) with capabilities to solve difficult problems with higher accuracy than its predecessors, was launched in 2023. Reportedly trained on 1 trillion parameters, the model cost over $100 million to train.

Although much improved than GPT-3.5, GPT-4 is known to have hallucination issues and lacks explainability. According to a May 2024 NIH study, GPT-4’s hallucination rate stood at 28.6%.

Its successor, GPT-4o (o for omni), OpenAI’s flagship model that can process and generate audio, visual and text content in real time, announced last year, will remain the native model for ChatGPT.

GPT-4o demonstrates enhanced capabilities in conversations and reasoning. “In head‑to‑head evaluations it consistently surpasses GPT‑4 in writing, coding, STEM, and more,” said the note.

Late in March, OpenAI announced that improvements were made to GPT-4o, following which the model “feels more intuitive, creative, and collaborative, with enhanced instruction-following, smarter coding capabilities, and a clearer communication style.”

“Early testers say that the model seems to better understand the implied intent behind their prompts, especially when it comes to creative and collaborative tasks,” the company wrote.

OpenAI said however that GPT-4 will continue to be available to users via its API.

In another news, The Verge reports that the company which hit a $300 billion valuation in March after a $40 billion funding round led by SoftBank, is already preparing to launch GPT-4.1 with a host of smaller models. According to the report, the GPT-4.1 family of models can debut as soon as next week.

Tibor Blaho, lead engineer at AIPRM, a prompt engineering firm for ChatGPT, weighed in in a post on X, “here is undeniable confirmation – model art for model pages on the OpenAI API Platform is already being prepared for o3, o4-mini, and GPT-4.1 (including nano and mini variants).”

TECHSTRONG TV

Click full-screen to enable volume control
Watch latest episodes and shows

Next Gen HPE ProLiant Compute Deep Dive

TECHSTRONG AI PODCAST

SHARE THIS STORY