18 comments

[ 2.6 ms ] story [ 23.9 ms ] thread
So DeepSeek v4 will cost 2x between 9:00 to 12:00 and 14:00 to 18:00, Beijing Time.

This is UTC+8 so to get cheaper DeepSeek:

- in London (BST / UTC+1): 11:00 to 02:00, and 05:00 to 07:00.

- in NY (EDT / UTC-4): 06:00 to 21:00, and 00:00 to 02:00.

- in California (PDT / UTC-7): 03:00 to 18:00, and 21:00 to 23:00.

Subscription plans like OpenCode Go might not even pass down this 2x cost since most customers are EU/US.

(comment deleted)
> in California (PDT / UTC-7): 03:00 to 18:00, and 21:00 to 23:00.

I think that’s incorrect. For California (PDT), 2X pricing will be in effect from 6–9pm and 11pm–3am. During PST, the CA windows will be 5–8pm and 10pm–2am.

I don't understand, you say that Deepseek will be more expensive for seven hours a day and then give five hours a day where it's cheaper in London.
That math makes no sense
>This is UTC+8 so to get cheaper DeepSeek:

Note "cheaper" is only relative to the 2x peak price. The off-peak price is still the same price as today. In other words they hiked the price 2x for peak hours only.

Does opencode go use deepseek infrastructure?

I thought they just used open weights models on infrastructure they get from someone else. They say,

“What about data and privacy?

The plan is designed primarily for international users, with models hosted in the US, EU, and Singapore for stable global access. Our providers follow a zero-retention policy and do not use your data for model training.”

https://opencode.ai/go

My impression is that they use OpenRouter, so market price among multiple inference providers.
Content restricted in the UK
For Berlin time (Germany, UTC+2, CEST), the 2x cost period is 03:00 - 06:00, and 08:00 - 12:00.

08:00 - 12:00 would have really hurt if prices were not ridiculously low to begin with.

I have not seen the official deepseek communication to this effect.
Yall trust LLM providers sweeping up keys and passwords? Sometimes you have a little .env file or config files in your repos that are not checked in... LLMs like to read that stuff and send it in. I don't trust any of the LLM providers, but less so the ones seated in China.
It's business at the end of the day. Don't get surprised if one day DeepSeek does a rug pull too.
This announcement also mentioned that they will release the next version (official non preview version) of v4 in mid July.
You were NOT the Chosen One! It was said that you would NOT destroy the Sith, BUT join them! Bring balance to the Sheets, not leave it in darkness!
So for some time spans it goes from ridiculously cheap to just very very cheap? I support that and I see no problem whatsoever.
A bit weird that it also applies to weekends. I guess it helps not creating wrong incentives for companies, but it hurts open and personal projects a bit.