I agree but then again it seems a model doing as good as the others at only 2% the cost would upend expectations as to return on investments from investors. Wait Why did ours cost so much more?
OpenAI o1 costs $15 per million input tokens and $60 per million output tokens, DeepSeek Reasoner, which is based on the R1 model, costs $0.55 per million input and $2.19 per million output tokens.
My bad I read it as the cost of creation, code and training, which was used in the papers that created the drop in market value.
As stated in your article, R1 was build with v3 from which we have no cost available.
Everybody involved at that level knows the AI hype is a sham, so they have their finger hovering above the “sell” button. Open AI’s valuation and cash burn, for example, assume a much higher revenue in the future – but if this upstart can release a similar product with a fraction of the overhead, that cash burn now becomes a massive liability.
I really dislike how jittery that market is.a fart in one direction sends ripples through it.
I agree but then again it seems a model doing as good as the others at only 2% the cost would upend expectations as to return on investments from investors. Wait Why did ours cost so much more?
It doesn’t. you can’t compare the cost of R1 to GPT 4.
Why not? Seems pretty straightforward
My bad I read it as the cost of creation, code and training, which was used in the papers that created the drop in market value. As stated in your article, R1 was build with v3 from which we have no cost available.
Everybody involved at that level knows the AI hype is a sham, so they have their finger hovering above the “sell” button. Open AI’s valuation and cash burn, for example, assume a much higher revenue in the future – but if this upstart can release a similar product with a fraction of the overhead, that cash burn now becomes a massive liability.
Edit to add a source from a real news outlet instead of shitty YouTube clickbait: https://www.bloomberg.com/opinion/articles/2025-01-27/china-s-ai-deepseek-shows-silicon-valley-s-huge-blindspot
that’s who bubbles tend to work