Is anyone still using DeepSeek? It’s an AI model developed in China, and I’m curious about what specific tasks it’s actually good at. How does it compare to other AI tools you’ve tried?
DeepSeek
Started by Ziah [3994906] on in Technology.
About this thread
Posts archived: 2 / 2 posts (100%) · the total is Torn's reply count + the opening post at the last fetch
Counted by TornLife from the archived posts.
- Archived posts
- 2
- Discussion span
- →
- People posting
- 2
- Likes on archived posts
- 1
- Authority score
- 7 / 100
- Historical score
- 16 / 100
- Story score
- 26 / 100
- Engagement score
- 44 / 100
i've used quite a few, deepseek is nice but it isn't specifically good at any single task. its a general model that its major claim to fame is that it is comparable to some of the other state of the art models at a fraction of the price.
one in particular that i really love, is gpt-oss-120b. i've been running that for most of my queries because it is so god damn cheap and i love how verbose it is. any question you ask it, it'll generate about 10-30k tokens and about 15 different tables of comparison on whatever it is you asked.
in terms of the latest models going, the xiaomi model, mimo v2 flash is really nice for a small model when you turn thinking off.
GLM 4.7 is also another which is nice at agentic stuff.
the chinese models are pretty interesting because they are all coming with their own flavour of chinese price innovation and optimisation rather than pushing the boundaries. kind of interesting how it mirrors other industries.
a problem for a lot of people is that they don't really know when or how they should be doing something. a lot of basic prompts actually perform better with the non-thinking mini models than they do with the state of the art ones with high reasoning, but people are so convinced that higher number = better that they don't bother to check.