Hacker News Reader

Comment by ChadNauseam | Hacker News Reader

Claude is simply much better. Additionally, since it’s possible to train on synthetic data from other LLMs, it’s easier to catch up to competitors than it is to surpass them

View on HN

> "Claude is simply much better."

Until lots of free AI models close the gap again. It is simply not good enough for such valuations like this for Anthropic.

Any other valuation higher than 70B for Anthropic is really overvalued and such higher valuations are always defended by those who invested in it already.

A few friends and I run Deepseek (and other LLMs) on some hardware we jointly own and offer endpoints to each other via VPN. Deepseek is a big step forward for open LLMs and is something I use when writing scripts that process lots of data because we've all split the capital to build the machine, but Claude was and still is a lot better than Deepseek in the kinds of day-to-day usage that I'm after. And with how cheap Gemini 2.0 is, I find myself considering whether to use my local LLMs even for data analysis.

I realize that on benchmarks it appears R1 is very close to Sonnet 3.7, but in practice I use Sonnet 3.7 much more.

I would say “for now” and that Anthropic’s inability to execute (3.7 was only released in response to competition from DeepSeek) raises a lot of questions about their future. While I agree that DeepSeek is still behind, Anthropic also does not have the same caliber of ambitious talent that DeepSeek has and has become a magnet for safety-obsessed cranks, whom (in total) act as a drag on the company. That Dario has also been engaging in dull-minded saber rattling about China does not speak well about their ability to deliver (vs. posture) in the future.

> has become a magnet for safety-obsessed cranks

I'm out of the loop. To whom are you referring?

Reply on news.ycombinator.com