thomas

Netflix but for LLMs.

I pay $20/month, access whatever LLMs I want at any given time (Chat GPT, Claude Sonnet, Perplexity, etc) through a unified prompt interface, and each one receives a share of my $20 prorated to how much I used them.

This allows for more dynamic competition and rewards innovative releases more fluidly than the current standalone subscription model.

Who’s building this?

PhD in aviation ∩ climate science | Aviator, diver, ETH maxi, nexialist | Host of /diving /ethfinance /eth-staker /geopolitics /science /singapore /thomas

Sapiens qui prospicit - Wannabe Polymath. Interested in art, music, history, philosophy and religion.

Thanks, hadn’t heard about it

2000 $DEGEN

OpenRouter 
It’s prepaid, but has 50+ models and all major models (OpenAI, anthropic, llama,…)

fortune cookie casts, memes, parenting, building frames, /framedl & /framedl-pro

I’d say the trouble started when catalog owners pulled *out* of Netflix-the-aggregator, forcing it to pivot to a Netflix-the-producer business model to keep offering enough content to survive.

It wasn’t Netflix’s core competency to be producing content, and that directly translates into the quality issues you’re referring to

I don’t expect nor want that LLM aggregator to start training their own model, just to make the best models available at a flat rate

if this will drive llm quality the same way netflix drives movie quality then please no

Does Poe.com meet ur needs? Or u are looking at more of an LLM API aggregator

Building @nanograph @cast2save, moding /singapore

Very much an aggregator. I’m paying for Chat GPT-4o right now but would like to shift to Claude, and sometimes Perplexity depending on the need. I can’t do that well without paying multiple subscriptions

co-founder @survey x /ponder // benadamsky.com

Other comments prob better suited to what you’re looking for, but raycast ai is $20/mo and comes with a ton of the latest models and is super easy to switch between

I have seen https://withmartian.com/products/model-router which seems to fit the bill. Gonna try it out on some personal projects soon.

Reminds me of https://nat.dev/compare (without the fancy prorated stuff)

I feel like interfaces are starting to intentionally diverge a bit with multimodal, teams, memory, etc?

Wanna be the very best. Building @moxiescout. ゼニゼニ

TBH I don’t care much about a unified interface. What I care about is a single flat monthly payment giving me access to multiple LLMs.

Netflix first gave us access to all movie catalogs for a flat fee, and now every catalog owner launched their own competing streaming platform, so you need to sub to 3–4 of them to cover the same range.

I don’t want to pay 3–4 subscriptions to LLMs, or even a single sub to one LLM when the “best in class” title keeps shifting every month (like Claude Sonnet is eating GPT’s lunch right now)

Awesome! But only the free versions I assume?

I had made something like this in the EVM discord last year, but just supported gpt, chatgpt, bard, dalle, and stable diffusion. I'm pretty sure I still have the code. It did have the logic to maintain and reset conversations, etc.

In Perplexity PRO, you can switch between models – is this what you look for?

Embracing the Mysteries, Unveiling the Realities | My blog: mariusschober.com

Interesting - so presumably Perplexity pays others for usage? It's only $20, which is as much as a single GPT-4o sub for instance