Benchmark Database

Benchmark results with evaluation method, model version, source type (vendor-reported vs independent) and date. Scores are data points — never universal rankings.

Agentic and Web Benchmarks

10 results · verified 2026-11-05

GAIA, BrowseComp, AutomationBench, CyberGym and IFEval: what each measures about agents, which Chinese models publish results, and why harne…

Chinese Language Benchmarks

0 results · verified 2026-11-05

C-Eval, CMMLU and SuperCLUE explained: what Chinese-language benchmarks measure, why contamination limits them, and why no tracked model pub…

Coding and Terminal Benchmarks

20 results · verified 2026-11-05

SWE-bench, Terminal-Bench, DeepSWE and LiveCodeBench compared: what each measures about coding models, the scores Chinese models publish, an…

Frontier Reasoning Benchmarks

11 results · verified 2026-11-05

HLE, GPQA Diamond, AIME and ARC-AGI in one place: what each measures, the scores Chinese models publish, and why the four are not interchang…

Multimodal Benchmarks

4 results · verified 2026-11-05

MMMU-Pro and Video-MME compared: what vision and video scores actually measure for Chinese models, the results published, and why resolution…