PostTrainBench data: US-China model gap of ~8% is nothing unusual

maksym_andr · x · 2026-11-20

Responding to claims that Chinese AI progress is stalling, fullrank checked PostTrainBench data: Qwen4 Max scored 36.7%, just 0.1 point below GLM 5.3 Flash (36.8%) — not an anomaly. Charting the historical trend of Chinese vs. American models, the current 8% gap has appeared repeatedly before and falls within normal fluctuation.

Original post →

More from AGI Musings

AGI Musings channel →