TaxPraBen - TaxSCQ: leaderboard
TaxPraBen (2026) is a Chinese tax-practice benchmark spanning tax knowledge memorization, understanding and application, evaluated zero-shot. This board is accuracy in percent on single-choice tax exam questions; higher is better.
Metric: Accuracy (%) on TaxSCQ (single-choice tax exam questions, 700 items) in TaxPraBen, zero-shot, Chinese prompts written and rated by tax-domain annotators; higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 19 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | internlm2.5-7B (TaxPraBen checkpoint unspecified) | 60.7 | |
| 2 | ERNIE-3.5 | 58.9 | |
| 3 | GPT-4o | 55.6 | #751 |
| 4 | Grok 3 | 54.1 | #611 |
| 5 | TaxPraBen GLM-4-9B (checkpoint unspecified) | 46 | |
| 6 | GPT-3.5 Turbo | 43.6 | #3005 |
| 7 | Qwen 2.5 7B (TaxPraBen checkpoint unspecified) | 40.7 | |
| 8 | YAYI2-30B | 34.6 | |
| 9 | Baichuan2 (TaxPraBen checkpoint unspecified) | 32.3 | |
| 10 | Yi-6B (TaxPraBen checkpoint unspecified) | 30.7 | |
| 11 | chatglm3-6B | 28.3 | #8281 |
| 12 | Atom (TaxPraBen checkpoint unspecified) | 26.4 | |
| 13 | TaxPraBen DeepSeekllm (checkpoint unspecified) | 25.9 | |
| 14 | LLaMA3 (TaxPraBen checkpoint unspecified) | 23.1 | |
| 15 | DeepSeek-R1-Distill-Qwen-7B | 23 | #8055 |
No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=taxpraben-taxscq · How It Works · Data refreshed daily, snapshot 2026-11-21.