Benchmark checker Chinese and Asian

Is SuperCLUE still meaningful?

Still useful Checked against the report, 5 November 2026

Yes for Chinese-market ranking, since private questions resist contamination, but scores from different rounds do not compare.

Why still useful: SuperCLUE refreshes its private questions each round.^{[1][1] DataLearner, SuperCLUE page Secondary} It ran a 22-model round in March 2026 and a later round in autumn.^{[2][2] AIBase on SuperCLUE March 2026 Secondary}^{[1][1] DataLearner, SuperCLUE page Secondary}

What “Still useful” means

Known problems

Scores from different rounds do not compare, because the questions change each round.^{[1][1] DataLearner, SuperCLUE page Secondary}
Coverage of the March 2026 round gives no scores for the top two, Claude-Opus-4.6 and Gemini-3.1-Pro.^{[2][2] AIBase on SuperCLUE March 2026 Secondary}
The report could not read the live board, which renders in JavaScript. Its scores come from AIBase and DataLearner.^{[2][2] AIBase on SuperCLUE March 2026 Secondary}^{[1][1] DataLearner, SuperCLUE page Secondary}
None of nine Chinese flagship cards reports SuperCLUE scores (5 Nov 2026).^{[4][4] DeepSeek-V4.1-Flash model card Primary}^{[5][5] GLM-5.3 model card Primary}^{[6][6] ERNIE-5.5-Flash-Base model card Primary}^{[7][7] Qwen3.8-Flash-Next-FP8 model card Primary}^{[8][8] DeepSeek-V4-Pro-0813 model card Primary}^{[9][9] Kimi-K3 model card Primary}^{[10][10] Kimi-K3.1-Pro model card Primary}^{[11][11] MiniMax-M3.1-Flash model card Primary}

What to look at instead

The report names no direct successor.

Where it appears in the report

Chapter 6, China and Asia

China has its own leaderboards, but frontier claims run on Western tests
Figure 2. On China's Own Leaderboard, Chinese Models Trail by Less Than a Point in “Go deeper: Chart: SuperCLUE scores, 2026 rounds”
Table 3. China's three leaderboard operators in “Go deeper: The three leaderboard operators and xbench”
Chinese model cards report agentic scores and no safety evaluations

Sources

Copied from the chapter source lists. Each number links back to the source in its chapter.

[1]
DataLearner, SuperCLUE page. It describes the SuperCLUE method and lists later scores. Secondary datalearner.com Chapter 6, source 16
[2]
AIBase on SuperCLUE March 2026. It reports the March 2026 SuperCLUE ranking. Secondary aibase.com Chapter 6, source 19
[3]
SuperCLUE, arXiv 2307.15020. It describes the SuperCLUE benchmark at launch. Primary arxiv.org Chapter 6, source 30
[4]
DeepSeek-V4.1-Flash model card. It gives base and chat benchmark tables and a harness-sensitivity table. Primary huggingface.co Chapter 6, source 4
[5]
GLM-5.3 model card. It reports Z.ai's benchmarks against Western and Chinese rivals. Primary huggingface.co Chapter 6, source 5
[6]
ERNIE-5.5-Flash-Base model card. It reports Baidu's base-model scores, including CMMLU. Primary huggingface.co Chapter 6, source 20
[7]
Qwen3.8-Flash-Next-FP8 model card. It reports Alibaba's benchmarks and its best-of-harness rule. Primary huggingface.co Chapter 6, source 34
[8]
DeepSeek-V4-Pro-0813 model card. It reports DeepSeek's own benchmarks and harness. Primary huggingface.co Chapter 6, source 35
[9]
Kimi-K3 model card. It reports Moonshot's own and third-party benchmarks. Primary huggingface.co Chapter 6, source 36
[10]
Kimi-K3.1-Pro model card. It reports Moonshot's latest benchmark tables. Primary huggingface.co Chapter 6, source 37
[11]
MiniMax-M3.1-Flash model card. It reports MiniMax's benchmarks, including an in-house reproduction. Primary huggingface.co Chapter 6, source 38