GPT-4.5 is capable of handling various file formats, including PDF and TIFF, but its ability to process these files equally well depends on several factors.
General File Handling Capabilities
GPT-4.5 supports all file types, which means it can accept and process PDFs and TIFFs, among others[1]. However, the model's effectiveness in extracting and utilizing information from these files can vary based on the file's structure and content.
PDF Handling
GPT-4.5 can process PDFs, which is an improvement over earlier models that required additional plugins for PDF handling[4]. However, PDFs can be challenging due to their layout and formatting, which might not always be easily parsed by the model. Users have noted that PDFs can be unreliable and hard to read for GPTs, suggesting that converting them into more structured formats like Markdown might improve performance[2].
TIFF Handling
While GPT-4.5 supports image inputs, including formats like TIFF, its primary strength lies in text-based processing rather than image analysis[1]. It can generate SVG images and use DALL·E for image generation based on text prompts, but it does not produce multimodal outputs like audio or video[1][5]. The model's ability to "understand" or analyze TIFF images directly is limited compared to its text processing capabilities.
Comparison of File Formats
In general, structured data formats like JSON or CSV are often preferred for feeding knowledge to GPTs because they provide clear, organized information that the model can interpret effectively[2]. Plain text files are also efficient for quick parsing, while formats like PDFs and TIFFs may require additional processing steps to extract meaningful data.
Conclusion
GPT-4.5 can handle PDFs and TIFFs, but its effectiveness varies based on the file type and structure. For optimal performance, converting files into more structured or text-based formats might be beneficial. While GPT-4.5 supports a wide range of file types, its strengths lie in text-based tasks rather than complex image analysis.
Citations:
[1] https://help.openai.com/en/articles/10658365-gpt-4-5-in-chatgpt
[2] https://community.openai.com/t/gpts-best-file-format-for-knowledge-to-feed-gpts/497368
[3] https://cdn.openai.com/gpt-4-5-system-card.pdf
[4] https://www.reddit.com/r/OpenAI/comments/17it40r/gpt4_can_now_process_pdfs_and_various_other_files/
[5] https://blog.promptlayer.com/everything-we-know-openais-gpt-4-5-model/
[6] https://community.openai.com/t/what-are-the-limitations-of-gpt-4-in-analyzing-pdf-text/534760
[7] https://venturebeat.com/ai/openai-releases-gpt-4-5/
[8] https://www.frontiersin.org/journals/sustainable-cities/sections/urban-economics/for-authors/author-guidelines