5 ms·
> But Cursor and other such tools depend almost entirely on accessing Anthropic, OpenAI and Gemini models, until open-source open-weight and in-house models mat
by jgalt212 11mo ago
> But Cursor and other such tools depend almost entirely on accessing Anthropic, OpenAI and Gemini models, until open-source open-weight and in-house models match or exceed frontier models in quality.
I'm not sure I agree with this because even though Cursor is pay north of 100% of revenues to Athropic, Anthropic is selling inference at a loss. So if Cursor builds and hosts its own models it still has the marginal costs > marginal revenues problem.
The way out for Cursor could be a self-hosted much smaller model that focuses on code, and not the world. This could have inference costs lower than marginal revenues.
- bogzz 11mo agoI suppose supermaven is d'oing something to that effect.
- xnx 11mo agoCan you have a useful code model that doesn't understand the world? It seems like such a model would be limited to little more than auto complete.
- simianwords 11mo agowhat makes you so sure of this? > Anthropic is selling inference at a loss. cost of models have gone down dramatically over time.
- jgalt212 11mo agoTrue, but they keep using more tokens (Agentic FTW!) and the currently most expensive models. source: https://www.wheresyoured.at/ https://www.wheresyoured.at/
- simianwords 11mo agothis is not true. they use more tokens to get more performance - the cost of using the model is going up but the performance is going up with it.
- jgalt212 11mo agoYour observation does not align with recent Cursor price changes.
- simianwords 11mo agoI’m talking about metered billing and not subscription.
- vrighter 11mo agocost goes down by 50% (it didn't) but you fake improvement while lacking any real improvement by taking the average of 4 runs. Cost just doubled, even though inference technically halved. Never mind the fact that the current model is alwayj outdated, and a new (bigger one) one is always being trained in parallel with the supposedly "cheaper" inference.