Not sure whether to choose ChatGPT, Claude, or something else for my schoolwork? (Need large input/output tokens for PPT summaries)

Hello! I'm a Biology major currently using the free versions of ChatGPT (occasionally getting access to GPT-4o before it reverts to another model) and Claude (3.5 Sonnet is currently the free version). I don't use these for coding - I mainly feed them large PowerPoint presentations + large 20 page textbook chapters from my courses to create simplified summaries while retaining key information. This helps me prepare outlines before lectures, so I can focus on adding new information from the professor during class.

I'm facing two main issues:

1. Limited input size: I can't input large texts like a 20,000-word book chapter all at once. Breaking content into smaller chunks works as a solution, so its not a huge deal.
2. Restricted output length: The summaries seem too condensed. For instance, when I input two 5-page sections and ask for a combined summary, I get about 3 pages back. It should be more like 5-9 pages. In contrast, when I use MistralAI, it generates much longer outputs without limitation.

These limitations have me considering a paid subscription to either ChatGPT (so I can get o1) or Claude. I'm wondering if services like ChatGPT Plus with GPT o1 would allow for larger outputs.

Additional considerations:

• I need something sophisticated enough for college-level biology. While benchmarks show Claude excels at coding, that's not relevant to my use case.
• I'd prefer the ability to input audio (like lecture recordings) and get summaries in 10-minute increments. I believe only GPT o1 offer this feature, though I currently work around this by using Whisper to convert audio to text before feeding it to GPT/Claude.

I'm also curious about alternative options. Platforms like Poe.com and You.com offer multiple LLMs (Claude 3.5 + GPT-4 Turbo + Llama, etc.) for roughly the same price as a single ChatGPT or Claude subscription. However, I've noticed their API implementations might not match the quality of the original services. Would subscribing directly through OpenAI or Anthropic provide better token limits and output sizes?