A new, advanced model named “o1” has been discovered within ChatGPT, offering capabilities beyond current versions. Users can access this model by specifying “o1” in their queries, which appears to support image understanding and reasoning. The model boasts a significantly larger context window of up to 200,000 tokens compared to the standard 32,000 tokens available to users.
This expanded context enables the handling of larger inputs and maintaining coherence over longer exchanges. However, access to this model has been temporarily restricted following its initial discovery. Users have reported being able to process file uploads, triggering an intensive processing mode.
The “o1” model’s advanced capabilities suggest a significant leap forward in OpenAI’s AI capabilities. Its use of a different stop token, similar to other advanced models, may indicate it is part of a new generation. Unofficial announcements hint that OpenAI may be preparing to release new features that improve the AI’s reasoning and processing abilities.
These features were uncovered by users experimenting with ChatGPT’s capabilities and model parameters. However, an official timeline for these features’ implementation remains unclear, leaving questions about when or if “o1” will become publicly available.
Source: https://www.testingcatalog.com/hidden-o1-model-in-chatgpt-boasts-200k-token-context-surpassing-current-versions