Categories Companies Research

Loading the SOTA2 catalog…

Explore

Categories Companies Research

Contribute

Join SOTA2 Contact

SOTA2

© 2026 SOTA2

Build in Bellevue, USA

WizWand is now SOTA2 Research

Papers Benchmarks Datasets Tasks

Language Modeling research benchmarks – Page 28 · SOTA2 Research

Language Modeling

Benchmarks

Dataset NameSOTA MethodSort

Dataset NameSOTA methodMetricTrendResultsLast Updated

WikiText-2 Top 5% most uncertain tokens (val)

5.13NLL

5

Feb 26, 2026

WikiText-2 All tokens (test)

1.66NLL

5

Feb 26, 2026

WikiText-2 All tokens (val)

1.7NLL

5

Feb 26, 2026

1.808Loss

5

Feb 26, 2026

BabyLM STRICT 100M words (test)

82.8BLIMP

5

Feb 26, 2026

BabyLM STRICT-SMALL 10M words (test)

75.8BLIMP Score

5

Feb 26, 2026

German Tail Entity (test)

0.295WER

5

Feb 26, 2026

German Speech-to-Text (STT) (test)

0.063WER

5

Feb 26, 2026

German Voice Assistant (VA) (test)

5.25WER

5

Feb 26, 2026

English Voice Assistant (VA) (test)

3.99WER

5

Feb 26, 2026

English STT (test)

3.56WER

5

Feb 26, 2026

93.04Accuracy

5

Feb 27, 2026

One Billion Word corpus 1M sentences

71.8Perplexity

5

Feb 26, 2026

One Billion Word corpus 1M sentences (test)

71.8PPL

5

Feb 26, 2026

Google Billion Word (test)

26.67Perplexity

5

Feb 26, 2026

Russian (RU) (test)

206Perplexity

5

Feb 26, 2026

French (FR) (test)

159Perplexity

5

Feb 26, 2026

Spanish (ES) (test)

145Perplexity

5

Feb 26, 2026

German DE (test)

206Perplexity

5

Feb 26, 2026

Czech (CS) (test)

312Perplexity

5

Feb 26, 2026

0.749Bits/Byte

4

Jul 8, 2026

Paloma-C4 (val)

Partially Reverse BDLM Mamba-H

58.48Perplexity

4

Jul 7, 2026

OpenWebText (dev)

18.32Perplexity (Dev)

4

Jul 1, 2026

TinyShakespeare (val)

1.689Validation Loss

4

Jun 23, 2026

1.275Loss

4

Jun 23, 2026

Page 28 of 37

Popular datasets

WikiText-103 WikiText-2 C4 LAMBADA WikiText

Follow for updates

Join SOTA2 Contact