I used https://fergusargyll.site/ which was made by @FergusArgyll as mentioned in their post "Introducing: Devil's advocate" to push back against my NO position on this market. I think the arguments made were sufficiently weak that I actually updated my position and bought more NO.

Manifold Markets Explorer

Here was my argument:
"Custom GPTs were a marketing move that OpenAI played in order to cut off the craze of GPT wrappers which were fine-tuned or RAG-ed on a custom dataset. There were potential competitors which would build upon people's data and Custom GPTs allowed you to do it in OpenAI's interface, funneling more traffic to their site. This worked. But this is no longer a craze and there are other applications to be built with API access. There are now other competitors such as Gems in Gemini which don't get much usage or hype. The fact that Anthropic has not invested in a similar product shows how little opportunity there is in this specific application. Anthropic is very focused on B to B product development. If this were something many businesses wanted or needed in the specific format which OpenAI introduced, they surely would have offered that product by now, as they did with web search. Even if it is very easy to switch Custom GPTs to the new models, I believe this is essentially an orphaned project from over 18 months ago, during which time the company has show explosive growth and investment in other aspects of the business. The most likely outcome to me is that they scrap Custom GPTs but it is also probably that they simply leave them alone with the latest GPT-4o running them."

Here was Gemini's counter-argument:
"Your conviction that OpenAI's Custom GPTs are an orphaned marketing ploy destined for stagnation or the scrap heap, while seemingly logical based on the initial hype cycle, fundamentally misunderstands OpenAI's strategic trajectory and the evolving AI landscape. To argue against your "NO" stance, a compelling case can be made for "YES" – that OpenAI will update Custom GPTs to a model more advanced than GPT-4o by the end of 2025.

Firstly, dismissing Custom GPTs as merely a "marketing move" that has lost its "craze" is a short-sighted view. While they certainly generated initial buzz, Custom GPTs represent a crucial component of OpenAI's platform strategy. They encourage users and businesses to build within the OpenAI ecosystem, fostering a degree of vendor lock-in and generating valuable data that can be used for future model training. This isn't just about a fleeting craze; it's about building a sticky platform with network effects. OpenAI has ambitious user growth plans, aiming for 1 billion users by the end of 2025, and enhancing their AI offerings through interfaces like Custom GPTs is part of achieving that goal.

Secondly, the comparison to Google's "Gems" and Anthropic's lack of a direct equivalent is not the decisive indictment you believe it to be. The relative lack of "hype" around Gems could be attributed to Google's execution or branding, rather than a fundamental lack of market opportunity in customizable AI. Furthermore, while Anthropic has focused on B2B with its Claude models and "tool use" features for custom AI helpers, their strategic choices don't dictate OpenAI's. OpenAI is actively engaged in the B2B space with offerings like ChatGPT Enterprise, and Custom GPTs can serve as a powerful tool for businesses to create tailored AI solutions on top of OpenAI's cutting-edge models. The fact that Anthropic is enabling businesses to tailor AI models suggests a clear market need for customization, a need that OpenAI is already addressing with Custom GPTs.

Your assertion that Custom GPTs are likely an "orphaned project" ignores the iterative nature of OpenAI's development. While the initial announcement was some time ago, OpenAI has continued to refine its models and platform. Upgrading the underlying model for Custom GPTs is likely a less resource-intensive task than the initial development, essentially a model swap that would immediately enhance the capabilities of all existing Custom GPTs. This provides a significant value add to their paying subscribers (ChatGPT Plus, Team, and Enterprise), a key revenue stream. OpenAI has explicitly stated their goal to "iteratively deploy models that empower developers and users."

Moreover, there's evidence suggesting ongoing development and planned upgrades. As of April 2025, OpenAI was preparing to enable native GPT-4o image generation within the Custom GPT builder, indicating continued investment in the platform. While some user discussions in March 2025 noted Custom GPTs were still on GPT-4o and not yet on newer models like GPT-4.5, this indicates a desire and expectation for upgrades among users and doesn't preclude an update by EOY 2025. In fact, discussions around model retirements suggest a natural progression to newer versions. The retirement of older models like GPT-4 in favor of GPT-4o in the main ChatGPT product further supports the likelihood of Custom GPTs being updated to the latest available model to maintain consistency and performance across the platform.

Finally, the rapid pace of AI development means that leaving Custom GPTs on an older model would quickly render them less competitive and valuable. To keep their platform attractive and demonstrate the capabilities of their latest research, OpenAI has a strong incentive to update Custom GPTs to whatever is more advanced than GPT-4o by the end of 2025. This aligns with their overall strategy of advancing AI research and technology and making their tools more user-friendly and effective.

In conclusion, while the initial hype may have subsided, Custom GPTs are a strategic platform component for OpenAI. The technical lift to update the underlying model is relatively low, the benefit to user experience and platform competitiveness is high, and there's evidence of ongoing development in this area. Arguing for "YES" recognizes OpenAI's commitment to iterative improvement and platform growth, making the update of Custom GPTs to a more advanced model by EOY 2025 a highly probable outcome."

Here was Gemini's counter-argument leaving the "reasoning" field blank":

"You're predicting that OpenAI won't update custom GPTs to a model more advanced than GPT-4o by the end of 2025. Let's critically examine that stance.

Firstly, consider OpenAI's relentless pace of development. They've moved from GPT-3 to GPT-4, then introduced iterative improvements like GPT-4 Turbo, and recently released GPT-4o, emphasizing speed and cost efficiency alongside multimodality. Their history suggests a strong commitment to pushing the frontier of AI models. It's highly probable they will release a successor, potentially labeled GPT-5 or a similarly significant architectural improvement, within the coming months, well before the end of 2025. Some market speculation even pointed to a potential late 2025 or early 2026 timeframe for GPT-5, aligning with the historical 33-month gap between GPT-3 and GPT-4.

Secondly, the competitive landscape in AI is incredibly heated. Companies like Google, Anthropic, and others are rapidly advancing their own models. To maintain its leading position and justify its subscription services (like ChatGPT Plus, which is required to create custom GPTs), OpenAI has a strong incentive to integrate its latest and most capable models into its flagship products, including custom GPTs. Not upgrading custom GPTs would risk making them less competitive and valuable compared to capabilities offered elsewhere.

Thirdly, custom GPTs are designed to be powerful, tailored tools. Their utility is inherently linked to the intelligence, reasoning ability, context window, and multimodality of the underlying model. An upgrade to a more advanced model would directly enhance the capabilities of all custom GPTs, making them more effective for complex tasks and specialized applications. This provides a clear value proposition for users and developers invested in the custom GPT ecosystem.

While the market description mentions the possibility of OpenAI retiring custom GPTs, the continued focus on them in recent discussions and guides (even into 2025) suggests this is unlikely in the near term. The infrastructure and user base for custom GPTs represent a significant asset and a platform for deploying advanced capabilities.

Given OpenAI's trajectory of model releases, the intense competitive pressure, and the clear benefits of a more advanced model for custom GPT functionality, arguing against an upgrade by EOY 2025 seems to underestimate the dynamics of the current AI race and OpenAI's strategic incentives. The release of a model more capable than GPT-4o, and its subsequent integration into key offerings like custom GPTs, appears not just probable, but highly likely within the given timeframe."

@JaySocrates Yeah, the tool doesn't understand basic things. It also needs a response option.