Key Highlights:
- OpenAI’s latest model is positioning itself as the core engine for future products, including the next Codex upgrade.
- A leaked GitHub changes point to a new Codex model based on GPT-5.2, with no separate “Max” tier in sight.
- If Caribou manages improvement in performance, it could be the start of OpenAI’s first serious GPT-5.2 oush in production-grade coding workflows.
OpenAI is under immense pressure to catch up with the progress Google and Anthropic is making lately in terms of performance and accuracy of their models. That’s the main reason why it released the latest GPT-5.2 model after reports from last week hinted that Gemini 3 has caught the attention of Sam Altman since the launch. Well, OpenAI describes GPT-5.2 as its “most capable model series yet for professional knowledge work,” and that’s not an exaggeration to be honest. You can read more about it here.
Fast forward to today, a new report from Testing Catalog suggests that a new Codex model is in the works now. The report is based on recent GitHub changes, which have leaked a Codex model codenamed “Caribou.” With GPT-5.2 quietly setting expectations around general reasoning, instruction-following, and multi-step problem solving, we are mostly likely to see the next Codex based on it.
No separate “Max” tier not in sight as of now
Adding weight to the leaked information, OpenAI’s Codex repository now shows updates to the model selector that reference Caribou. It’s a strong sign that the next Codex release will be based on GPT-5.2. Notably, there are no signs of a separate “Max” tier model this time around. However, OpenAI appears to be leaning into the idea of offering the default experience for everyone rather than paywalling performance behind additional models.
If the report turns out to be accurate, Caribou might show up as a solid upgrade for coders out there. Following GPT-5.2 launch, OpenAI rolled out GPT Image 1.5 and has referenced multiple internal upgrades in close succession. That suggests OpenAI is now optimizing around a stable core model and extending it horizontally across products, rather than chasing one-off launches.
Competition is intensifying
All that said, the competition is intense right now in the AI space. So, if Codex manages to brings in any kind of measurable performance increase, Anthropic and Google would jump right in. But for now, GPT-5.2 looks like the foundation OpenAI plans to stand on in 2026.
If “Caribou” lands as expected, it won’t just be a Codex update. It’ll be the first clear sign of how far GPT-5.2 can stretch when applied to serious, production-grade developer work.
