Dear OpenAI Team,
I’d like to raise an urgent concern from a developer’s perspective - one I believe is shared by many within the technical community.
In recent weeks, the efficiency of the current ChatGPT models, particularly for software development, has significantly declined. Compared to earlier versions like (“o1”) or (“o3-mini”), all currently available variants (including GPT-4-turbo or “o3/o4-mini-high”) feel heavily restricted - in some cases even unusable for precise and reliable engineering work.
One concrete example: a model suggested replacing a "-" with a "+" in a code snippet - a logically and syntactically incorrect recommendation that introduced functional errors. Fixing such “corrections” manually costs valuable time - time developers often don’t have, especially when dealing with complex or security-critical systems.
My suggestion:
Please consider a clear separation of models going forward:
- A “mainstream model” for general use - casual chat, education, creative writing, etc.
- A “pro model” for developers, technical users, researchers, and other demanding tasks - focused on precision, efficiency, and without artificial limitations on code, syntax, or logic.
Many of us would gladly pay more for such a “developer-grade” model - as long as it restores (or surpasses) the quality, accuracy, and reliability we experienced in earlier versions.
My concrete question:
Are there any upcoming updates, a potential restructuring of the model lineup, or at least some indication of when a high-performance, technically capable model might return?
The current direction feels like a step backward. It slows down real work and makes it harder to use OpenAI tools in professional settings. I truly hope the needs of your technical user base will be taken seriously and not ignored - even if the current focus leans more toward image generation or creative outputs.
Please take this feedback seriously - many developers support your mission and want to keep building with your tools. But we need tools that empower us, not slow us down.
Thank you in advance for your response.
A developer from the field.
