In early December, the source “Jimmy Apples” hinted on X about a potential GPT-4.5 release by the end of the month. Recent user reports suggest that ChatGPT now identifies itself as “gpt-4.5-turbo.” However, inquiries about the differences between “gpt-4.5-turbo” and “gpt-4.0-turbo” yield only vague answers based on April 2023 knowledge, possibly indicating a misperception.
OpenAI typically releases ‘turbo’ versions well after the initial model, as seen with GPT-3.5 and GPT-4, making an immediate GPT-4.5 turbo version an anomaly.
A mid-December screenshot suggested GPT-4.5 would support audio, video, 3D, text, and images, with enhanced logic problem-solving and cross-modal capabilities. However, a missing arrow in the screenshot’s link raises doubts about its authenticity.
Previously, GPT-4-Turbo was prematurely disclosed via a website’s pricing table, later deemed fake by OpenAI CEO Sam Altman. This doesn’t negate the possibility of GPT-4.5 Turbo’s existence, aligning with ongoing speculations about OpenAI’s advancements in AI.
Multimodal Capabilities
In March, Microsoft Germany’s CTO Andreas Braun mentioned OpenAI’s focus on multimodal models capable of video generation. This aligns with the “Project Arrakis” leak, promising video generation and improved performance, possibly feeding into Project Gobi, a prototype for future models like GPT-4.5 or GPT-5, potentially launching in early 2024.
OpenAI has already ventured into 3D model generation with Point-E and Shap-E, indicating a trend of continuous innovation, especially in light of competition with Google’s Gemini project. Microsoft’s response to Gemini Ultra with enhanced GPT-4 benchmarks further underscores this competitive landscape.
Read related articles: