$ETH
DeepSeek has upgraded again
It’s all about being both cheap and capable
In the AI race,
it’s no longer about who burns the most cash
but who knows how to save it
Guys,
there’s another new development in the AI world
DeepSeek launched V4.1-Flash in September
It’s faster, more efficient, and more capable
It also adds native visual understanding
Its reasoning, AI agent, and coding abilities have all been upgraded
It’s also better at 3D generation
and supports “wish-based” programming:
just describe what you want in natural language to generate a complete program
Meanwhile, reasoning costs have dropped dramatically
The official introduction said
the new architecture can handle tasks with fewer active parameters
while also reducing cache usage
The goal is simple:
let AI do more work
without increasing the cost
In other words, more for the same price
A lot of people used to think
that whoever spent more and bought more chips
would stay ahead
But if models keep getting more efficient
and the same computing power can do more work,
then the advantage companies spent a fortune to build
will have to be reassessed
For ordinary people like us,
what makes an AI good?
A good AI is useful and affordable
Am I right?
So, you see,
a lot of tech companies
make a big show of themselves,
piling on hardware and chips
and looking very high-tech
But damn, costs just keep going up
and the experience for ordinary users stays the same
That’s just another way to fleece people
When it comes to AI,
no one gets my respect like Liang Wenfeng
He gets the best results
at the lowest cost
Anyway, I think
future AI shouldn’t just be about parameters or computing power
What matters most is performance, cost, and real-world applications
Guys,
what do you think?
#deepseek发布更新版ai模型
DeepSeek has upgraded again
It’s all about being both cheap and capable
In the AI race,
it’s no longer about who burns the most cash
but who knows how to save it
Guys,
there’s another new development in the AI world
DeepSeek launched V4.1-Flash in September
It’s faster, more efficient, and more capable
It also adds native visual understanding
Its reasoning, AI agent, and coding abilities have all been upgraded
It’s also better at 3D generation
and supports “wish-based” programming:
just describe what you want in natural language to generate a complete program
Meanwhile, reasoning costs have dropped dramatically
The official introduction said
the new architecture can handle tasks with fewer active parameters
while also reducing cache usage
The goal is simple:
let AI do more work
without increasing the cost
In other words, more for the same price
A lot of people used to think
that whoever spent more and bought more chips
would stay ahead
But if models keep getting more efficient
and the same computing power can do more work,
then the advantage companies spent a fortune to build
will have to be reassessed
For ordinary people like us,
what makes an AI good?
A good AI is useful and affordable
Am I right?
So, you see,
a lot of tech companies
make a big show of themselves,
piling on hardware and chips
and looking very high-tech
But damn, costs just keep going up
and the experience for ordinary users stays the same
That’s just another way to fleece people
When it comes to AI,
no one gets my respect like Liang Wenfeng
He gets the best results
at the lowest cost
Anyway, I think
future AI shouldn’t just be about parameters or computing power
What matters most is performance, cost, and real-world applications
Guys,
what do you think?
#deepseek发布更新版ai模型