Artificial Intelligence thread

[magmunta]

China ranks first globally for two consecutive weeks! The number of calls to China's AI large-scale models reaches 4.69 trillion tokens.

Please, Log in or Register to view URLs content!

That's because more ordinary Chinese use openclaw and LLM-based automations, even though a qualitative gap between the top-notch American LLMs and the Chinese LLMs has been increasing since Kimi's last update. China has got quantity , but lacks quality.

 

[tphuang]

That's because more ordinary Chinese use openclaw and LLM-based automations, even though a qualitative gap between the top-notch American LLMs and the Chinese LLMs has been increasing since Kimi's last update. China has got quantity , but lacks quality.

???? that's nonsense. Have you ever heard of seed 2.0?

 

[magmunta]

???? that's nonsense. Have you ever heard of seed 2.0?

I said "a qualitative gap between the top-notch American LLMs and the Chinese LLMs has been increasing since Kimi's last update" and it's true. Seedance 2.0 is the frontier in the video generation business, but it is not LLM-based.

 

[meedicx]

I said "a qualitative gap between the top-notch American LLMs and the Chinese LLMs has been increasing since Kimi's last update" and it's true. Seedance 2.0 is the frontier in the video generation business, but it is not LLM-based.

You shouldn't be commenting if you don't know the difference between Seed, Seedance and Seedream models.

After the latest update, I would rank Doubao the best free chatbot app in user experience, especially using Chinese and voice input. Doubao thinking mode is much better than Gemini/Grok (free ChatGPT is absolutely terrible so don't need to be compared). Only paid Kimi 2.5 is better in advanced use cases since you can get it to run python and generate charts.

Doubao will remain free/subscriptionless and monetize through ads/shopping. Bytedance is much more competent than OpenAI in this area (their recent efforts have flopped according to reports)

 

[tphuang]

If you want to see where China is in terms of customer/end user experience with AI, you have to try Doubao app.

Of course, you would need a Chinese phone to do it. Thankfully, I have one.

 

[MortyandRick]

I said "a qualitative gap between the top-notch American LLMs and the Chinese LLMs has been increasing since Kimi's last update" and it's true. Seedance 2.0 is the frontier in the video generation business, but it is not LLM-based.

That's because more ordinary Chinese use openclaw and LLM-based automations, even though a qualitative gap between the top-notch American LLMs and the Chinese LLMs has been increasing since Kimi's last update. China has got quantity , but lacks quality.

Bold statement, even if coming from someone who is so unabashedly pro US.

Please show us evidence that the qualitative gap has been increasing since Kimi latest update?

And that the increase in usage is due to ordinary Chinese using open claw?

Do you have evidence to prove that's the case? Put up or shut up is what we say in our business.

 

[Randomuser]

Bold statement, even if coming from someone who is so unabashedly pro US.

Please show us evidence that the qualitative gap has been increasing since Kimi latest update?

And that the increase in usage is due to ordinary Chinese using open claw?

Do you have evidence to prove that's the case? Put up or shut up is what we say in our business.

It's just another troll that takes sides against whatever the consensus of this forum. You can see consistent behaviour in the other threads.

I hope this thread really limits this sort of behaviour coz I come here for development news not to see these low quality winology discussions.

 

[Andy1974]

Please, Log in or Register to view URLs content!

Incredibly, China now holds all 5 top spots on open router daily LLM rankings, each model from a different provider.

 

[Eventine]

Western LLMs are definitely still ahead on agentic coding, but I wouldn't say the gap has been increasing. There's been diminishing returns at the very top level in the last couple of months, the Claude 4.5 -> 4.6 and GPT 5.2 -> 5.4 difference isn't as drastic as the previous jumps.

As much as we might hate Anthropic's CEO, Claude is still undeniably the best at enterprise-level coding performance in my testing and talking to various folks in the industry, but Anthropic's decision to ban & go after Chinese enterprise users is going to screw them in the long-term. China is a huge market for LLMs and the tokens usage reflect that since Western apps/models are banned. All of that usage then feeds into additional training data, which with sufficient compute, will eventually put China ahead even in areas the West is currently ahead in.

 

[FriedRiceNSpice]

Western LLMs are definitely still ahead on agentic coding, but I wouldn't say the gap has been increasing. There's been diminishing returns at the very top level in the last couple of months, the Claude 4.5 -> 4.6 and GPT 5.2 -> 5.4 difference isn't as drastic as the previous jumps.

Frankly, if agentic coding continues to advance then the overall impact of economy will be negative. Millions of high paying upper middle class jobs will be eliminated. Consumer spending will collapse and the knock on effects for the economy will be disastrous.