Hi there,
Here is this week’s Hot or Not.
Not: AI costs
The cost of tokens (the basic units of text that AI models process, similar to word chunks) has dropped sharply.
OpenAI’s token cost has seen an over 10x decrease since 2023. CEO Sam Altman blogged recently that a 10x cost improvement is likely to continue every 12 months.
Fresh competition will put pressure on pricing to come down even more quickly.
DeepSeek’s Reasoner API pricing is nearly 30x cheaper than OpenAI’s for its o1 model.
The Chinese company announced this week it would cut pricing by another 75% during off-peak hours to align with US/EU daytime hours.
In our recent conversations with DeepSeek customers, cost was highlighted as a key advantage.
"It’s far cheaper. The cost is definitely a huge incentive…What they provide, as far as the cost of serving the models, is incredibly enticing. Overall, as much as OpenAI is reliable and a good model that they serve, they are very expensive and this is always a factor." — CEO at AI development startup
But…API downtime was not.
"There's obviously hesitancy on the reliability as well. After DeepSeek’s announcement, their API — if you look at their status page, everything was just red. We've got a backup in the sense where if DeepSeek fails, it'll flip over to o1. I'm hoping that they can strengthen their infrastructure, bring down the inference time, and keep pricing where they are at." — Chief Data Officer at Data analytics startup
CB Insights customers can check out our fresh interviews with companies exploring DeepSeek’s models here.
.png?upscale=true&width=1040&upscale=true&name=customer%20sentiment%20(DeepSeek).png)
