Open leaderboard: China's open-weight models trail US closed models by just 7 months, from 22,113 benchmark results
jakecoolguy · reddit · 2026-10-14
whatmodel.sh, an open-source AI leaderboard, charted the model race from 22,113 published benchmark results (987 benchmarks, 1,011 sources) and finds China's open-weight models are only about 7 months behind US closed models.
Details:
- Text-model subset covers 1,211 models on 617 benchmarks (18,755 results); 14,436 are independent evaluations (papers, Epoch AI, LMArena, EvalPlus, EleutherAI harness), 7,677 are lab-reported and labeled as such.
- Dataset is CC BY-SA 4.0, scoring code MIT, with CSV downloads and an interactive chart tracing the gap from GPT-3 davinci vs T5 Large in June 2020 (1+ month) to today.
More from Research
- Fingerprinting Leaderboard benchmarks model provenance as copying takes minutes vs millions to train — omarsar0 · 2026-10-14
- Is today's agent memory stack just a workaround for model architecture? — Sun-Tzu-Maybe · 2026-10-14
- Persona experiments on Talkie, a 13B LLM trained only on pre-1931 text — dhadfieldmenell · 2026-10-14
- HKU and Tencent open-source Multics, a multi-player world model that trains on 4 players but scales to 10 — Scobleizer · 2026-10-14
- SAP paper teaches LLM agents to predict their own failure (AUROC 0.85) — SAP · 2026-10-14
- CUHK's Orchestra robot framework completes 7.35-step subtask sequences — CUOHK · 2026-10-14