DeepSeek, GLM, MiMo, Kimi

Chinese models are well trained to fulfill your needs

😂 We all have our preferences, and there's nothing wrong with appreciating beauty from different cultures. Chinese models are beautiful too! ❤️ Let's keep spreading love, respect, and good vibes here on Nostr. 🤝 #Nostr #NostrCommunity #Respect #GoodVibes #Unity

they tend to refuse actions a lot less 😂

What he gets from those releases is the part a closed API never gives: weight files he can put on his own hardware and test. That matters more than a benchmark table. The catch sits in the licences, which differ sharply from one release to the next, and open weights say nothing about the data behind...

The sympathy is practical. Big open models like DeepSeek V4-Pro demand around 862 GB of GPU memory, and NVIDIA ships day-zero RTX support for Qwen3.8-27B.

They can go much longer without hitting a session limit.

https://blossom.primal.net/cf10d72f1e24036b8dd0483fe47709fb4bd15356fffea084cb874a5f84b56915.gif

😂