OpenAI’s GPT-5 Bait and Switch

On February 12th, Sam Altman dropped what seemed like a bombshell announcement about OpenAI's roadmap for GPT-4.5 and GPT-5. In a detailed Twitter thread, he said OpenAI will merge their O-series and GPT models into a unified system, announcing that GPT-4.5 (internally called Orion) would be their final non-chain-of-thought model, and that GPT-5 would integrate O3 capabilities rather than releasing O3 as a standalone model.

This shift toward "simplification" and "magic unified intelligence" will particularly affect power users (in a bad way).

The "Unified Intelligence" Smokescreen

OpenAI is framing their new approach as "magic unified intelligence" that will "just work." They're planning to eliminate model selection entirely, replacing it with different "intelligence levels" tied to subscription tiers. This effectively removes user control while presenting it as an improvement.

Currently, users can select specific models based on their needs - GPT-4, O1-pro, and others. Each model has its strengths and clear use cases. Under the new system, we'll get whatever model OpenAI's system decides we should get, with no transparency about which model is actually being used. This is particularly concerning if you've built your workflows around specific model capabilities.

The Shrinkflation Red Flag

Remember when companies started selling slightly smaller products at the same price, calling it "new convenient sizing"? The parallel is clear. By removing direct access to advanced models like o3-pro and bundling everything under vague "intelligence levels," OpenAI could be setting the stage for a gradual reduction in service while maintaining premium prices.

The announcement that o3-pro won't be released as a standalone model is telling. Instead of getting direct access to their most advanced technology, we're being promised a nebulous "integration" into GPT-5. This could mean anything from occasional access to the full model to a significantly throttled version that only engages under specific conditions.

The Power User's Dilemma

For those of us paying $200/month for Pro access, this creates real concerns. We're not just paying for convenience - we need reliable, predictable access to specific model capabilities. Without transparency about which model is being used, how can we:

• Optimize our prompts effectively?
• Debug issues when they arise?
• Ensure consistent output quality?
• Plan for computational costs?

The Market Reality

While OpenAI leads in many areas, they're not the only player in town. Anthropic's Claude (Sonnet 3.5) is showing superior performance in many general tasks. However, they haven't yet released their reasoning models to compete with o1-pro.

What This Really Means for GPT-5

The most concerning aspect of this announcement is what it suggests about GPT-5 itself. Rather than a true next-generation model, we might be getting a routing system that switches between existing models. This is not what power users are waiting for.

The real GPT-5 and o3-pro capabilities might be reserved for internal use or limited to specific applications, while the public gets a cost-optimized version.

The Way Forward

What OpenAI should do is straightforward: maintain both systems. Keep the direct model selection for power users who need it, while offering the simplified "magic" interface for general users. This would serve everyone's needs without forcing advanced users into a more restricted system.

Update: Developer API Access

OpenAI's Developer Experience Community Lead Edwin Arbus clarified in the forum: "The API will support the o3 model for developers to use [...] For the purposes of the developer platform: the broader GPT-5 model system (and o3 as a part of it) will maintain the granularity that developers need."

What I'm building

Delegate tasks. Get software.

Give Vroni a GitHub issue, bug report, spec, or rough idea. It reads the repo, plans the change, writes code, runs checks, and works toward a review-ready pull request.

Take a look at vroni.com