OpenAI has launched GPT-4.5 (Orion), its largest AI model yet, boasting improved world knowledge, emotional intelligence, and creativity. While OpenAI presents this as a major advancement, the reception has been skeptical due to high costs and underwhelming performance compared to top AI reasoning models like DeepSeek R1 and Claude 3.7 Sonnet.
- Extreme Costs: GPT-4.5 is very expensive to run ($75–$150 per million tokens), making its long-term viability uncertain.
- Performance Concerns: It falls short of leading reasoning models in complex tasks, showing that scaling alone is no longer delivering big performance leaps.
- GPU Shortages: OpenAI appears to be running low on GPUs, signaling potential turbulence ahead, especially as demand for AI computing surges.
- And… Nvidia CEO Jensen Huang says that next-generation AI will need 100x more compute than older models because of new reasoning approaches.
- Competition: Meta will launch and AI standalone app.
Despite OpenAI’s optimism, rising costs, competition from more efficient models, and compute constraints could spell trouble. If the company can’t sustain scaling or access enough GPUs, it may struggle against rivals like Anthropic and DeepSeek, forcing a shift in strategy before GPT-5 launches later this year.