AI model
Dense 32-billion-parameter open-weight language model from Alibaba's Qwen3 family, released in 2025 under Apache 2.0 with switchable reasoning modes.
Independent researcher Nick Cammarata, who says he left OpenAI in 2023, published a report arguing almost the entire AI network is interpretable. Business Insider calls the findings still theoretical, so safety teams have a claim to test before they have an audit method.
Reality
Goodfire's Nick Cammarata says the vast majority of neurons in models up to trillions of parameters are legible, reversing an explanation he helped shape. Google DeepMind's Neel Nanda warns against overhyping it, and the model-level numbers are preliminary.
Reality
GPT-4.1 scored 93.49% on BrowseComp-Plus with documents supplied and 14.58% when BM25 searched, according to The New Stack. That 78.9-point gap sizes the room for retrieval work on one fixed corpus. Production failure rates need their own measurement.
Reality
Aleph Alpha released Kolibri, an Apache 2.0 German-English model that activates 3.46B of its 78.1B parameters per token. Each token costs about as much compute as a small model, yet a team hosting it in Europe still has to fit every expert in memory.
Perspective Coverage
6 publishersReality
Researchers writing on LessWrong found only Nemotron 3 Super and Qwen3 32B changed refusals when test suspicion was removed from their reasoning traces. In most other models, test talk in a trace looks like general caution, so it is weak evidence of eval gaming.
Reality
A dev.to post scores ten coding models on five real tasks and divides by price. The method is cheap to copy; the vendor plumbing it recommends deserves more scrutiny than the arithmetic.
Reality
One engineer's home lab tally: $1,400 a month for two A100s running 40% idle, against open-weight models he measured inside noise of GPT-4o. The break-even is real, and it sits high.
Reality