[wastewastewaste]

What is the most capable LLM for summarising stuff like long educational articles perhaps?

[sebmellen]

Anthropic's Claude 2 is great for pasting large text files in, and Flan T5 seems to be excelling at summarization benchmarks: https://huggingface.co/docs/transformers/model_doc/flan-t5.

[dartos]

GPT-4 is usually the answer to “which llm is most/best…”

But if the article is longer than the context window, you can try either RWKV or Claud-200k

[simonw]

Has Claude 200k ever been made available? I've only found ways to access the 100k version.

[simonw]

GPT-4 if it's less than 8,000 tokens (or 32,000 if you have access to that via the API).

Claude for up to 100,000 tokens.

[wahnfrieden]

GPT-4