Has anyone tested Claude-instant-100k?

I'm fairly happy with GTP-4, but Poe released Claude with 100k tokens, see below.

(Anthropic’s fastest model, with an increased context window of 100k tokens (around 75,000 words). Enables analysis of very long documents, code, and more. Since this is an experimental early access model, usage is currently limited to 100 messages per month for Poe subscribers )

I wonder how that compares to GTP-4, if anyone has tried it and can share your impressions.

tex42 said:

I'm fairly happy with GTP-4, but Poe released Claude with 100k tokens, see below.

(Anthropic’s fastest model, with an increased context window of 100k tokens (around 75,000 words). Enables analysis of very long documents, code, and more. Since this is an experimental early access model, usage is currently limited to 100 messages per month for Poe subscribers )

I wonder how that compares to GTP-4, if anyone has tried it and can share your impressions.

Click to expand...

I've been paying for GPT plus for several months now and have gotten a lot of value out of it.

My most common workflow is recording voice memos, then transcribing it with this app called MacWhisper which uses the Whisper AI hosted locally. Then I feed that into chatGPT v4 and have it clean it up and do light editing.

I saw that with Poe you can get GPT4 AND Claude 100k for $20/month so today is day one.

I haven't fed it any huge documents yet, but I'm impressed with it's comprehension of what I fed into it. Although I wish I could get it to generate longer outputs. For some reason it seems like the outputs are really short, even though the default options it gives show a strong comprehension of what I fed into it.

It just seems like it might be being purposefully crippled so as to not use as many tokens on the way out? Not sure, but I'm going to do my best to find out as much as I can in the next week before the trial ends.