Are you seeking smarter insights delivered directly to your inbox? Sign up for our weekly newsletters to receive the most relevant information tailored for leaders in enterprise AI, data, and security. Subscribe now!
The Current State of AI Innovation
The evolution of artificial intelligence is at a critical juncture, much like the advent of freeways in the United States after 1956, a vision brought to life by President Dwight D. Eisenhower’s administration. While high-performance vehicles like Porsche, BMW, and Ferrari have existed for decades, the infrastructure necessary for their optimal use was only established later. Similarly, while AI models are becoming increasingly advanced and capable, the essential infrastructure to facilitate genuine, real-world innovation is still in development.
Arun Chandrasekaran, a distinguished VP analyst at Gartner, highlighted this analogy in a conversation with VentureBeat, stating, “All we have done is create some very good engines for a car, and we are getting super excited, as if we have this fully functional highway system in place.” This situation has led to a plateau in model capabilities, exemplified by OpenAI’s GPT-5. Although it represents a significant advancement, it still only hints at the potential for truly autonomous AI.
Challenges in AI Scaling
The landscape of enterprise AI is being reshaped by power limitations, increasing token costs, and delays in inference. Join our exclusive salon to learn how leading teams are:
– Transforming energy into a strategic advantage
– Designing efficient inference for real throughput gains
– Unlocking competitive ROI with sustainable AI systems
Secure your spot to stay ahead: [https://bit.ly/4mwGngO](https://bit.ly/4mwGngO)
Chandrasekaran noted, “It is a very capable model, it is a very versatile model, it has made some very good progress in specific domains.” However, he emphasized that this progress is more incremental than revolutionary, especially considering the high expectations set by OpenAI in the past.
Advancements in GPT-5
Despite its limitations, GPT-5 has made notable strides, particularly in coding tasks and multi-modal capabilities. Chandrasekaran pointed out that OpenAI has strategically focused on making GPT-5 “very good” at coding, recognizing the vast opportunities within enterprise software engineering and aiming to compete with Anthropic’s leadership in that domain. Additionally, GPT-5’s advancements in modalities beyond text, especially in speech and images, present new integration possibilities for enterprises.
The model also subtly enhances AI agent and orchestration design through improved tool usage; it can now call third-party APIs and handle multiple tasks simultaneously. However, this advancement necessitates that enterprise systems have the capacity to manage concurrent API requests in a single session.
Enhancements in Business Logic and Workflow
GPT-5’s ability for multistep planning allows more business logic to reside within the model itself, reducing reliance on external workflow engines. Its expanded context windows—8K for free users, 32K for Plus subscribers at $20 per month, and 128K for Pro users at $200 per month—can “reshape enterprise AI architecture patterns,” according to Chandrasekaran. This means that applications that previously depended on complex retrieval-augmented generation (RAG) pipelines can now directly process much larger datasets, streamlining workflows.
However, RAG is not rendered obsolete; as Chandrasekaran noted, “retrieving only the most relevant data is still faster and more cost-effective than always sending massive inputs.” Gartner anticipates a shift towards a hybrid approach with less stringent retrieval, allowing developers to utilize GPT-5 for managing “larger, messier contexts” while enhancing efficiency.
Cost Implications and Future Developments
On the cost front, GPT-5 significantly lowers API usage fees, with top-level costs at $1.25 per million input tokens and $10 per million output tokens. This pricing makes it competitive with models like Gemini 2.5 while significantly undercutting Claude Opus. However, the input/output price ratio of GPT-5 is higher than that of earlier models, which AI leaders should consider when evaluating GPT-5 for high-token-usage scenarios.
Ultimately, GPT-5 is designed to eventually replace GPT-4o and the o-series models. OpenAI plans to introduce three model sizes (pro, mini, nano), allowing architects to tier services based on cost and latency requirements. Simple queries can be managed by smaller models, while complex tasks will utilize the full model. Nevertheless, variations in output formats, memory, and function-calling behaviors may necessitate code reviews and adjustments. As GPT-5 may render some previous workarounds obsolete, developers are advised to audit their prompt templates and system instructions.
By phasing out previous versions, “I think what OpenAI is trying to do is abstract that level of complexity away from the user,” Chandrasekaran remarked. “Often we’re not the best people to make those decisions, and sometimes we may even make erroneous decisions, I would argue.”