[AINews] not much happened today
DeepSeek V4.1 Flash 1017 takes top open-weight spot on code and agent benchmarks at low cost.
DeepSeek released V4.1 Flash 1017, a 552B-total/16B-active MoE model with a 1M-token context window and MIT license, claiming the #1 open-weight position in both Code Arena WebDev and Agent Arena. The model's non-hallucination rate jumped dramatically from 4% to 33%, and Agent Arena notes it delivers 37% of GPT-6.1 Sol's improvement at 27% of the cost. While not yet in wide use, the value-per-dollar story makes it a credible forcing function for closed-model pricing.