Chinese SimpleQA
ERNIE 5.0 86% leads since Feb 2026 Fading 3,000 short Chinese factual questions across six topics. Graded correct, incorrect or not attempted.
GitHub repo moved from OpenStellarTeam to LivingFutureLab.
In the authors’ words
New LLM evaluation benchmarks are important to align with the rapid development of Large Language Models (LLMs). In this work, we present Chinese SimpleQA, the first comprehensive Chinese benchmark to evaluate the factuality ability of language models to answer short questions, and Chinese SimpleQA mainly has five properties (i.e., Chinese, Diverse, High-quality, Static, Easy-to-evaluate). Specifically, first, we focus on the Chinese language over 6 major topics with 99 diverse subtopics. Second, we conduct a comprehensive quality control process to achieve high-quality questions and answers, where the reference answers are static and cannot be changed over time. Third, following SimpleQA, the questions and answers are very short, and the grading process is easy-to-evaluate based on OpenAI API. Based on Chinese SimpleQA, we perform a comprehensive evaluation on the factuality abilities of existing LLMs. Finally, we hope that Chinese SimpleQA could guide the developers to better understand the Chinese factuality abilities of their models and facilitate the growth of foundation models.
- Level
- University Undergraduates and entry-level professionals
- Scored
- Graded by a model A language model scores each response, usually against a rubric.
- Answers
- Short answer A number, word or short string matched to a key.
ERNIE 5.0 leads with 86% among models released by Apr 2026
Best score per model as labs report them in their own cards. Labs pick their own settings, so these compare less cleanly than independent runs.
The newest model with a score here came out in Apr 2026, and nobody has reported a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:
- GPT-6 Astra
- Claude Opus 5.5
- GPT-6.1 Terra
- GPT-6.1 Sol
- and 16 more
- 1 ERNIE 5.0 86%
- 2
DeepSeek-V4-Pro 84.4% - 3 DeepSeek-V4-Flash 78.9%
- 4 ERNIE-4.5-300B-A47B 77.1%
- 5
Ling-2.6-1T 76.5% - 6 GLM-5 75.2%
- 7
Hunyuan-TurboS 70.8% - 8 GLM-4.5 70.1%
- 9 Hunyuan-T1 67.9%
- 10
MiniMax-Text-01 67.4% - 11 DeepSeek-V3 64.8%
- 12 DeepSeek-R1 63.7%
- 13 ERNIE-4.5-21B-A3B-Thinking 49.1%
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | ERNIE 5.0 | 86% | Feb 2026 | Model card | |
| 2 | DeepSeek-V4-Pro | 84.4% | Apr 2026 | Model card | |
| 3 | DeepSeek-V4-Flash | 78.9% | Apr 2026 | Model card | |
| 4 | ERNIE-4.5-300B-A47B | 77.1% | Jun 2025 | Model card | |
| 5 | Ling-2.6-1T | 76.5% | Jun 2026 | Model card | |
| 6 | GLM-5 | 75.2% | Feb 2026 | Model card | |
| 7 | Hunyuan-TurboS | 70.8% | Feb 2025 | Model card | |
| 8 | GLM-4.5 | 70.1% | Aug 2025 | Model card | |
| 9 | Hunyuan-T1 | 67.9% | Mar 2025 | Model card | |
| 10 | MiniMax-Text-01 | 67.4% | Jan 2025 | Model card | |
| 11 | DeepSeek-V3 | 64.8% | Dec 2024 | Model card | |
| 12 | DeepSeek-R1 | 63.7% | Jan 2025 | Model card | |
| 13 | ERNIE-4.5-21B-A3B-Thinking | 49.1% | Sep 2025 | Model card |
The benchmark's own leaderboard has 41 entries, 27 of them models we track. We plot every source in the charts below.
Number one over time
Lab claims count from the day the card came out, independent runs from the model’s release. We draw one row each for the benchmark’s own leaderboard and what labs claim in their cards.
The best score passed 85% in Feb 2026
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Labs headline it less often now, last in Apr 2026
Launch posts from every lab that has listed it, one tick per post. Tall colored ticks include it, and short gray ones leave it out.
| Date | Launch post | Score | As the post writes it |
|---|---|---|---|
| 24 Apr 2026 | DeepSeek-V4-Pro | 84.4 | Chinese-SimpleQA (Pass@1) |
| 6 Feb 2026 | ERNIE 5.0 | 86 | ChineseSimpleQA |
| 9 Sep 2025 | ERNIE-4.5-21B-A3B-Thinking | 49.1 | ChineseSimpleQA (Accuracy) |
| 30 Jun 2025 | ERNIE-4.5-300B-A47B | 77.1 | ChineseSimpleQA |
| 21 Mar 2025 | Hunyuan-T1 | 67.9 | C-SimpleQA |
| 16 Mar 2025 | ERNIE 4.5 | Chinese SimpleQA | |
| 27 Feb 2025 | Hunyuan-TurboS | 70.8 | Chinese-SimpleQA |
| 20 Jan 2025 | DeepSeek-R1 | 63.7 | C-SimpleQA (Correct) |
| 15 Jan 2025 | MiniMax-Text-01 | 67.4 | C-SimpleQA |
| 26 Dec 2024 | DeepSeek-V3 | 64.8 | C-SimpleQA (Correct) |
10 system cards report it, from 5 labs
Besides launch posts, we read the system cards, model cards and technical reports labs publish with a release. They report far more, including safety and frontier-risk tests.
| Date | Document | Score | As the document writes it | Section |
|---|---|---|---|---|
| 13 Jun 2026 | Ling-2.6-1T technical report | 76.5 | C-SimpleQA, correct (Table 5) | Capabilities |
| 26 Apr 2026 | DeepSeek-V4-Pro technical report | 84.4 | Chinese-SimpleQA (Pass@1), DS-V4-Pro Max (Table 6) | Hallucination |
| 26 Apr 2026 | DeepSeek-V4-Flash technical report | 78.9 | Chinese-SimpleQA (Pass@1), DeepSeek-V4-Flash Max (Table 7) | Hallucination |
| 17 Feb 2026 | GLM-5 technical report | 75.2 | Chinese SimpleQA, World Knowledge panel (Figure 11) | Hallucination |
| 4 Feb 2026 | ERNIE 5.0 technical report | 86 | ChineseSimpleQA (Table 2) | Capabilities |
| 8 Aug 2025 | GLM-4.5 technical report | 70.1 | GLM-4.5-Base, Chinese-SimpleQA (EM) (Table 2) | Hallucination |
| 29 Jun 2025 | ERNIE-4.5-300B-A47B technical report | 77.1 | ChineseSimpleQA (Table 5) | Hallucination |
| 20 Jan 2025 | DeepSeek-R1 technical report | 63.7 | C-SimpleQA (Correct), DeepSeek-R1 (Table 8, same as final column of Table 3) | Hallucination |
| 14 Jan 2025 | MiniMax-Text-01 technical report | 67.4 | C-SimpleQA (Table 8) | Hallucination |
| 26 Dec 2024 | DeepSeek-V3 technical report | 64.8 | C-SimpleQA (Correct), DeepSeek-V3 (Table 6) | Hallucination |
More in Factuality
- SimpleVQA In use Models answer 2,025 short factual questions about images across 9 tasks and 9 topics. An LLM judge grades answers against fixed references.
Qwen3.7-Plus 81.7% - HallusionBench In use Models answer 1,129 yes or no questions about 346 images, built in pairs to expose language hallucination and visual illusion. Scored by accuracy.
Keye-VL-2.0-30B-A3B 73.5% - SimpleQA Verified In use 1,000 human-verified, de-duplicated short-answer factual questions measuring parametric knowledge.
Gemini 3.1 Pro 77.5% - GaRAGe In use Models write long-form RAG answers to 2,366 questions with human-annotated grounding passages. Scoring checks whether answers use the relevant passages.
Nemotron 3.5 Bolt E4B 40.7% - Kimi Hallu Bench In use Models answer hard prompts drawn from practical work tasks. Fact-checking agents score the share of claims that are hallucinated. Kimi K3.1 Pro 8.30%
- RAGTruth In use A word-level annotated corpus of hallucinations in retrieval-augmented generation answers. It is used to train and evaluate hallucination detection and RAG faithfulness. Nemotron 3.5 Bolt E4B 77.9%