For context, I’m already a ChatGPT Plus subscriber and a very heavy ChatGPT user.
My feedback is not coming from someone who is unwilling to pay for OpenAI products. I happily pay for ChatGPT because I get substantial value from it. However, for coding agents and API-driven workflows, I strongly prefer pay-as-you-go pricing. My coding usage varies significantly, and metered billing is a much better fit for how I work than another subscription.
I primarily access OpenAI models through OpenRouter, which makes it very easy to compare OpenAI models directly against competing models from other providers in the same workflow.
I use Kilo Code as my coding agent, which allows me to switch between models quickly and evaluate them in the same environment.
GPT-5.5 is the model I would most like to use in Kilo Code. In my experience, it delivers the best balance of coding ability, reasoning quality, and reliability among the OpenAI models I’ve tried.
The problem is cost.
Because I’m paying out of pocket and not using the API commercially, GPT-5.5 is difficult for me to justify as my default coding model. OpenAI currently prices GPT-5.5 at $5 per million input tokens and $30 per million output tokens, while GPT-5.3 Codex is priced at $1.75 per million input tokens and $14 per million output tokens.
As a result, I’m often forced to use GPT-5.3 Codex instead of GPT-5.5, even when I believe GPT-5.5 would produce a better result.
I also find that GPT-5.4 Mini’s lower price does not always translate into lower real-world cost. For coding work, additional mistakes, hallucinations, and corrective prompting can sometimes offset a significant portion of the expected savings.
Because I use OpenRouter, I can compare models side-by-side. The current pricing frequently pushes me toward competing models such as Claude, Gemini, and DeepSeek—not because I prefer them, but because they fit my budget more comfortably.
This means the competition for GPT-5.5 is not just other OpenAI models. In practice, GPT-5.5 is competing against GPT-5.3 Codex, Claude, Gemini, DeepSeek, and other lower-cost models available through the same interface.
My feedback is simple:
I am not avoiding GPT-5.5 because of its quality. I am avoiding it because of its price.
As someone who already pays for ChatGPT Plus and prefers OpenAI’s models, I would like to use GPT-5.5 far more often than I currently do. Today, the pricing frequently forces me onto GPT-5.3 Codex or competing models, even though GPT-5.5 is usually the model I would prefer to choose.
If GPT-5.5 were more affordable for hobbyists and individual developers, it would become my primary coding model almost immediately.