ChatGPT users find hidden o1 model featuring image input and deep context capacity

Recently, users on social media reported discovering a new model named "o1" (non-preview) within ChatGPT. Some accessed this model by specifying "o1" in queries, which appears to offer capabilities beyond current versions. Notably, "o1" reportedly supports image understanding and reasoning, processing visual inputs.

The model is said to have a context window of up to 200,000 tokens, much larger than the standard 32,000 tokens available to ChatGPT Plus users. This expanded context could allow the handling of larger inputs and maintaining coherence over longer exchanges.

Some users found that "o1" was accessible temporarily and could process file uploads, triggering an intensive processing mode. However, access seems to have been restricted shortly after these discoveries.

The "o1" model's ability to solve complex problems suggests a considerable advancement in OpenAI's AI capabilities. It uses a different stop token, 200002, similar to other advanced models like "4o" and "4o-mini," indicating it might be part of a new generation.

Without official announcements from OpenAI, these leaks hint that the company may be preparing to release new features that improve the AI's reasoning and processing abilities. This could benefit users needing advanced computational tasks, such as data analysis or complex problem-solving.

These features were uncovered by users experimenting with ChatGPT's capabilities and model parameters. By accessing raw responses and modifying the model slug, they discovered this unannounced model.

There is no official timeline for these features' implementation, so it's unclear when or if "o1" will become publicly available. It remains to be seen how OpenAI will integrate these advancements into their platform and whether they will be accessible through the ChatGPT interface or only via the API.

Source: Reddit