Alibaba hikes DeepSeek V4 Flash 0731 output prices by 230%

GloVin 25 points 26 comments October 21, 2026
www.stork.ai · View on Hacker News

Discussion Highlights (11 comments)

xvxvx

‘The sudden cost spike exposes the fragility of building businesses on subsidized API access. If your margins depend on cheap tokens, you are entirely at the mercy of the labs.’ I’m just here for the fallout. The businesses who jettisoned their talent in favor of cheap AI garbage. The ones who have seen obvious degradation in work quality and stability, but who have tried to ignore or hide it. They now have to face down CFO’s and executive management who see zero returns on the current spend and are being asked to approve extreme budget increases. A 230% increase isn’t even a big jump. Try 1000%. That may not even be enough to cover the costs and turn a profit for AI companies.

ipsum2

AI blogspam.

eur0pa

I'd like to see the prompt that generated this piece edit: these are not the prices I see at https://api-docs.deepseek.com/quick_start/pricing/

bargainbin

I mean, even with a hike of 230% this is going to be lower than Luna/Haiku right?

blakeman8192

The article failed (deliberately?) to mention that 0731 vision exp is two releases old. It's not nearly as efficient or as capable as the newer 4.1 Flash and 4.1 Flash 1017 models. This move is most likely the DeepSeek team just making efforts to phase it out.

Dban1

So reliant on AI

throwa356262

Big deal, they are bringing 0731 prices to the same level as 1017. Given their compute constraints (thanks to US sanctions) it makes sense to not support too many legacy models anyway. https://www.alibabacloud.com/help/en/model-studio/model-pric...

roenxi

Over 12 months I'd expect [0] the $/task of completing a benchmarked task to drop to x0.01 of its original price. If providers increase their prices by 230%, that implies I would expect x0.033 the initial benchmark. I don't know if this even counts as newsworthy, prices are too volatile. Speaking of, wasn't this model released a few months ago? It must be getting close to being obsolete by now. [0] https://artificialanalysis.ai/trends#efficiency

miohtama

In other news, China is about to ban open weights model exports https://www.straitstimes.com/business/china-plans-ai-export-...

impure

V4 Flash 0731? Not even 4.1? And anyways, DeepSeek is an open model, plenty of services provide access to it. Also looking at pricing on OpenRouter it looks like Alibaba is actually one of the more expensive providers.

drcongo

Does "loss leader" mean something different here?

Semantic search powered by Rivestack pgvector
10,209 stories · 96,238 chunks indexed