Qwen 3.8 Max reportedly coded alone for 16 straight days with no human intervention.
Alibaba gave the 2.4-trillion parameter AI a single instruction, a blank folder and left it to write, test, debug and refine its own code until it produced working software. Pajeet jobs killer!!
Qwen 3.8 Flash Next (125B A6B MoE) Vs Qwen 3.8 27B (dense)
Holy shit that is some of the best Pagoda I've seen generated locally and that's by a 4-bit quant on a…
Holy shit that is some of the best Pagoda I've seen generated locally and that's by a 4-bit quant on a single 3090 (Qwen 3.8 Flash Next) - input + output 54k tokens. One shot:
Yesterday we saw GLM-5.3-Flash AND Qwen-3.8-Flash dropped, then we heard NVIDIA was buying HuggingFace…
What could be next to help close out this HOT Summer of AI?…. Grok 4.7?
You have reached the end of the archive
All of qwen38