Tax agencies, banks and airlines still depend on ageing code that few dare to modify. OpenAI says its latest models have made substantial gains on this kind of legacy code, and The Verge examines whether AI could help.
Nace AI has applied its Drex decision models to document search for AI agents. The company reports a 92.05% score on BrowseComp-Plus using 12.6 search calls on average, and fewer agent turns on OfficeQA Pro.
OpenAI still leads Arena's image rankings. Behind it, Nano Banana 2.1, Krea 3 Ultra and an improved Grok Imagine 2.0 are shaking up the order.
A study by New York University and the University of Texas at Austin finds that frontier language models overuse linguistic patterns that people struggle to parse. The authors link this to training on code and math, which rewards correctness over clarity; smaller models write in a more human way.
Nanbeige LLM Lab has released Nanbeige4.3-33B, a 33-billion-parameter dense model under Apache 2.0 with a 512K context window. It scores 80.9 on Terminal-Bench 2.1 and will be used to teach the forthcoming Nanbeige4.3-3B.
Hugging Face's Carbon-X is an open 14.8-billion-parameter DNA language model built on Qwen3-14B. It reads raw DNA in a chat and answers in English, averages 66.7% across five DNA benchmarks, and is released under Apache 2.0.