Artificial Intelligence thread

[HighGround]

Having tried the new Kimi K2.5 and Qwen Max-Thinking models I can only say... the gap between Qwen and Kimi is getting larger, not smaller. Kimi is simply in another league. Qwen not only did poorly, it straight-up hallucinated on several of my questions.

K2 was already a good model but it is now clearly great with the 2.5 update. The reasoning performance is on par with Opus 4.5 and GPT5.2-high in my testing.

Of all the major Chinese labs, Moonshot is probably most deserving of being flooded with GPUs and investment. Too bad geopolitics will prevent it. The talent density of the team is astounding.

The worth of an AI isn't tied to their hallucination rate is all I'll say. Qwen's stack is much more important than Kimi.

 

[bsdnf]

My experience is: Gemini 3 Pro ≈ Kimi K2.5 > GLM 4.7 > V3.2 ≈ Qwen 3 Max Thinking. Qwen 3 Max Thinking has a significant improvement over Preview, but not enough.

 

[bsdnf]

the full, non-distilled version of Z-Image has been released.

Please, Log in or Register to view URLs content!

 

[tokenanalyst]

My experience is: Gemini 3 Pro ≈ Kimi K2.5 > GLM 4.7 > V3.2 ≈ Qwen 3 Max Thinking. Qwen 3 Max Thinking has a significant improvement over Preview, but not enough.

Qwen3 4b. That is a model that run in laptop and punch well above its weight.

 

[tphuang]

https://twitter.com/i/web/status/2016278151576629274

Meituan is coming out with possibly the most advanced power architecture using SST for its new AI data center in Zhangjiakou.

 

[tphuang]

I played around with Kimi 2.5 a little more today. It's pretty good. Better than the free version of Gemini 3 and DeepSeek 3.2. + a mile ahead of Qwen 3 max.

 

[bsdnf]

btw, that high-scoring TTS mode in QWEN 3 MAX is not yet available.