CNSL-bench - Text Descriptions: leaderboard

CNSL-bench (2026) evaluates multimodal LLMs on Chinese National Sign Language: each entry of the standardized National Common Sign Language Dictionary is posed as a four-way multiple-choice question from a single input modality. This board is accuracy in percent identifying the sign from the dictionary's textual description of its articulation; higher is better, and chance is 25.

Metric: Accuracy (%) identifying the sign's meaning from the dictionary's textual description of its articulation, four-way multiple choice over the 6,707 sign entries of the National Common Sign Language Dictionary (distractors sampled at random), official sampling settings per model, accuracy by exact match of the chosen option; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 19 models tracked. Updated 7 Oct 2026.

Top models

#ModelSettingScoreOverall rank
1GPT-589.64#178
2Gemini 2.5 ProMedium84.79#276
3Gemini 2.5 Flash79.95#463
4Qwen 3 VL Plus76.68#512
5Qwen-VL-Max72.62#1400
6Qwen-VL-Plus72.41#5818
7Qwen 3 VL 8B (Thinking)70.61#910
8GPT-4o69.03#751
9GLM-4.1V-9B (Thinking)68.24
10GPT-4o Mini67.57#1667
11InternVL3.5-8B67.53#1604
12Qwen 3 VL 8B Instruct67.06#1029
13Qwen 3 VL 2B Instruct62.83#2367
14Qwen 3 VL 2B (Thinking)61.29#2535
15Qwen 2.5 VL 3B Instruct60.07#2713

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=cnsl-bench-text-descriptions · How It Works · Data refreshed daily, snapshot 2026-11-21.