List your company

ResearchFormerly WizWand

Language Model Evaluation

Benchmarks

DatasetSOTA methodMetricTrendResultsLast updated

BenchPress short-context (test)

68.84Accuracy

131

May 12, 2026

Winogrande, ARC-C, ARC-E, Lambada, PIQA, Hellaswag, MMLU, IFEval, and GSM8K-CoT (Mixed standard 10-shot prompt)

80.39Accuracy

88

May 13, 2026

Open LLM Leaderboard v2 (test)

60.84BBH

47

May 12, 2026

ArxivRollBench 2026a

moonshotai/kimi-k2.6

70.8Valid Accuracy

42

May 26, 2026

60.85Knowledge Score

35

Jul 2, 2026

Qwen3-0.6B Evaluation Suite average

47.8Average Performance

24

May 7, 2026

69.86Average Score

21

Jun 4, 2026

Pooled tasks Table 5 Llama-3.1 3.3 (various)

Llama-3.3 70B Instruct

57.15Pooled Accuracy Estimate (γ̂)

21

Feb 26, 2026

SCALENET (Layer-wise)

0.273ROUGE-Lsum

14

Feb 26, 2026

72.8Average Score

12

Jun 23, 2026

Composer (2Mb-M-3A)

49.3DCLM Core Score

12

May 18, 2026

lm-eval-harness (test)

74.22MMLU

9

May 7, 2026

Downstream Tasks Evaluation Suite Math, Code, Law, Know., Reason., MMLU

Math Dense Expert

4.92Math Accuracy

9

Feb 26, 2026

DATE-LM MMLU, GSM8K, BBH

62.07MMLU Accuracy

7

Jun 1, 2026

Quality, Factuality, and Safety Evaluation Suite (test)

Self-Improving Pretraining

86.3Generation Quality Score

7

Feb 26, 2026

NLP Evaluation Suite (WG, PIQA, BoolQ, ARC-C, ARC-E, OBQA, HS, SciQ, LM, RTE)

60.14WG

6

Feb 26, 2026

Open LeaderBoard v2

0.655AS

5

May 27, 2026

1.3B LLM Leaderboard

36.4ARC

5

Feb 26, 2026

77.54AS

1

May 27, 2026