Chinese Open Models Close the Gap — Amid Distillation Accusations
Practitioners say GLM-5.2 now sits in genuine SOTA territory, even as claims circulate that frontier labs' own models are being distilled into open competitors.
The benchmark conversation has shifted from whether Chinese open models can compete to how exactly they got so good. @dev_null321 put GLM-5.2 directly alongside the frontier: 'All of these are SOTA models. If you take GLM 5.2 and use it with Hermes vs Opus 4.8 there is a major difference.' The framing matters — even when noting differences, the comparison itself treats GLM-5.2 as a peer rather than a follower.
Unlock the full briefing
Get every story in today's briefing, the full archive, and the daily AI intelligence brief.
All stories today
Full archive
Daily brief
Cancel anytime. Payments powered by Stripe.