DeepSeek R2.... Butt hurt coming for Open AI and Anthropic!

Coffee Lounge
,
1

Open AI and Anthropic going to feel the pain with leaked details of DeepSeek R2 Models pricing and abilities being shared online... 0.07 Cents per million input tokens and 0.27 Cents per million output tokens! How will the U.S (and Pwesidwent Twump respond to this?)... Clenched cheeks all round!

Please don't tell me about privacy and China I'm just not interested (no disrespect intended) but don't kid yourself that U,S operators are any less honest about what they do with your information either... It is what it is...

Just wonder what retaliation there is going to be from Wall Street, Altman, and Musk? About to get really messy I feel. Can't blame Nvidia this time either... Big shares sell off? OUCH!

:smirk:

7 Likes
2

Also, Chinese chip production tech isn't near USA/Taiwan yet, they're mainly held back by some aspects of the lithographic process - light sources and lenses. Once they solve these issues (and they will) their chip tech will catch up fast and potentially overtake USA efforts. Running faster chips with their DeepSeek R2 iteration will give it an exponential leap forward..

2 Likes
3

Absolutely amazing news! I’ve been telling fuck Anthropic ever since Gemini became relatively good, as my initial DeepSeek experience wasn’t great, but this was before the Wappler’s AI integration, especially the last one, I’m looking forward to swapping Claude/Gemini combo and give it another go!

Also, I wonder how long will it take for mega corporations like Apple to start running their own DeepSeek LLMs instead of wrapping everything in ChatGPT, given the amount of devices that are powered by Apple Intelligence, and them falling far behind the AI race, it would be a no brainer? A huge loss for OpenAI! Especially if the rumours were true and they were looking into acquiring Windsurf to battle close ties that Cursor and Anthropic has!

1 Like
4

Will have to try out. Gave up on deepseek as it was so slow

1 Like
5

We have been staunch fans of Qwen and DeepSeek over Anthropic and Open AI Models. Speed is not a factor as seems very fast at the web-level (bit slower at the API level). Accuracy and code is clean and concise. With little need for repeat prompting unlike Claude which by design seems to waste tokens repeatedly. Am looking forward to R2 being released and Altman being hit right where it hurts! Also have yet to hit a limit with DeepSeek even on the free tier (we did sign up very early and wonder if that has anything to do with it being an early adopter?).

1 Like