Benchmark Almanac

Taiwanese Mandarin Language Understanding

Claude 3 Opus 73.6% leads since March 2024
Not in model cards

Models answer questions from Taiwanese exams in traditional Chinese. The subjects range from school topics to professional qualifications.

Made by
Po-Heng Chen and colleagues
Released
March 2024
Kind
Knowledge in many languages
Level
University
Scored
Checked automatically
Answers
Multiple choice
Tasks
All public
Measures
Capability
Language
Chinese
In and out
Text → Text
Interaction
Single turn
Items from
Exams and contests
Refresh
One fixed set
Run by
Anyone
Strongest model tested
GPT-4 Turbo Apr 2024 #172 on the capability index
Last model added
April 2024
Scores refreshed
November 20th, 2026
24
Entries in its paper, 11 models we track

Claude 3 Opus leads with 73.6% among models released by April 2024 Claude 3 Opus leads with 73.6% among models released by April 2024

Best score per entry from the benchmark's paper, read on November 20th, 2026. A paper’s table doesn’t add newer models. Where a model ran with several settings or harnesses, we keep its best. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders. This view shows only models we track, each at its best.

The newest model with a score here came out in April 2024, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

  1. 1 Claude 3 Opus 73.6%
  2. 2 GPT-4 Turbo Apr 2024 70.4%
  3. 3 Yi-34B-Chat 68.5%
  4. 4 Qwen-14B 61.8%
  5. 5 Gemini 1.0 Pro 61.4%
  6. 6 Qwen-7B-Chat 54.8%
  7. 7 Claude Instant 54.4%
  8. 8 Breeze-7B-Instruct 51.3%
  9. 9 GPT-3.5 π 49.4%
  10. 10 Mixtral-8x7B-Instruct 46.4%
  11. 11 ChatGLM3-6B 46.4%
  12. 12 Claude 2 44.6%
  13. 13 Mistral-7B-Instruct-v0.1 44.3%
  14. 14 Taiwan-LLM-13B-Chat 42.9%
  15. 15 Baichuan2-13B-Chat 41.2%
  16. 16 Taiwan-LLM-7B-Chat 37.4%
  17. 17 falcon-40b-instruct 35.5%
  18. 18 OLMo-7B-Instruct 34.6%
  19. 19 Llama-2-13b-Chat 34.4%
  20. 20 Yi-6B-Chat 32.8%
  21. 21 Qwen-0.5B-Chat 31.6%
  22. 22 Llama-2-7b-chat 28%
  23. 23 Falcon-7b-instruct 25.7%
  24. 24 Gemma-7b-it 25.6%

Source: The Taiwanese Mandarin Language Understanding paper

#EntryScoreSettingDateSource
1 Claude 3 Opus 73.6% Feb 2024 Official leaderboard
2 GPT-4 Turbo Apr 2024 70.4% Apr 2024 Official leaderboard
3 Yi-34B-Chat 68.5% Official leaderboard
4 Qwen-14B 61.8% Sep 2023 Official leaderboard
5 Gemini 1.0 Pro 61.4% Dec 2023 Official leaderboard
6 Qwen-7B-Chat 54.8% Official leaderboard
7 Claude Instant 54.4% Aug 2023 Official leaderboard
8 Breeze-7B-Instruct 51.3% Official leaderboard
9 GPT-3.5 π 49.4% Official leaderboard
10 Mixtral-8x7B-Instruct 46.4% Official leaderboard
11 ChatGLM3-6B 46.4% Oct 2023 Official leaderboard
12 Claude 2 44.6% Jul 2023 Official leaderboard
13 Mistral-7B-Instruct-v0.1 44.3% Sep 2023 Official leaderboard
14 Taiwan-LLM-13B-Chat 42.9% Official leaderboard
15 Baichuan2-13B-Chat 41.2% Sep 2023 Official leaderboard
16 Taiwan-LLM-7B-Chat 37.4% Official leaderboard
17 falcon-40b-instruct 35.5% Official leaderboard
18 OLMo-7B-Instruct 34.6% Feb 2024 Official leaderboard
19 Llama-2-13b-Chat 34.4% Official leaderboard
20 Yi-6B-Chat 32.8% Nov 2023 Official leaderboard
21 Qwen-0.5B-Chat 31.6% Official leaderboard
22 Llama-2-7b-chat 28% Official leaderboard
23 Falcon-7b-instruct 25.7% Official leaderboard
24 Gemma-7b-it 25.6% Official leaderboard
#ModelScoreSettingDateSource
1 Claude 3 Opus 73.6% Feb 2024 Official leaderboard
2 GPT-4 Turbo Apr 2024 70.4% Apr 2024 Official leaderboard
3 Qwen-14B 61.8% Sep 2023 Official leaderboard
4 Gemini 1.0 Pro 61.4% Dec 2023 Official leaderboard
5 Claude Instant 54.4% Aug 2023 Official leaderboard
6 ChatGLM3-6B 46.4% Oct 2023 Official leaderboard
7 Claude 2 44.6% Jul 2023 Official leaderboard
8 Mistral-7B-Instruct-v0.1 44.3% Sep 2023 Official leaderboard
9 Baichuan2-13B-Chat 41.2% Sep 2023 Official leaderboard
10 OLMo-7B-Instruct 34.6% Feb 2024 Official leaderboard
11 Yi-6B-Chat 32.8% Nov 2023 Official leaderboard

Number one over time

Lab claims count from the day the card came out, independent runs from the model’s release.

Anthropic
Benchmark paper Not tested since Apr 2024

Source: The Taiwanese Mandarin Language Understanding paper

Scores over time

All the scores we have, one dot each, so a model tested at several settings appears more than once.

Plot with other benchmarks
In the benchmark’s paper Best so far
0% 20% 40% 60% 80% 100% 2024 Near the ceiling (85%) Claude 3 Opus Qwen-14B Claude Instant Claude 2 0% 20% 40% 60% 80% 100% 2024 Near the ceiling (85%) Claude 3 Opus Qwen-14B

Source: Official Taiwanese Mandarin Language Understanding leaderboard

No lab model card reports it yet

We read the headline results of 540 launch posts from 57 labs, and none of them list this benchmark.

More in Multilingual

All 545 in Multilingual
  • MMMLU In use MMLU professionally translated into 14 languages. Multiple-choice accuracy. Claude Mythos Preview 92.7%
  • WMT24++ In use English-to-55-language translation of WMT24 test data. Scored with automatic metrics such as MetricX or xCOMET. K-EXAONE 90.5%
  • MMLU-ProX In use MMLU-Pro translated into up to 29 languages with expert review. Multiple-choice accuracy per language. Qwen3.7-Max 87%
  • CLIcK In use Models answer 1,995 multiple-choice questions on Korean language and culture. Questions come from official Korean exams and textbooks in eleven categories. Solar Open 2 90.7%
  • Global MMLU Lite 3.0 In use Models answer 400 human-translated MMLU questions per language in 23 languages. Version 3.0 adds Czech, Hungarian, Oriya, Slovak and Tajik and updates Italian. Accuracy. Claude Opus 5 95.7%
  • MMMB In use Models answer multiple-choice questions about images in six languages across 15 categories. The set has about 12,000 questions. LFM2.5-VL-3B 83%

Data audit

9 of 10 checks pass, 1 warning. Look first at the Newest score check. The newest came 2 years ago, in April 2024, for GPT-4 Turbo. Every benchmark

Catalog record 2 pass
  • Pass Task access All tasks are public.
  • Pass Confidence Checked against a primary source.
Scores 1 warning 3 pass 1 note
  • Pass Independent runs Scores from the benchmark’s paper (24 entries).
  • Pass Models scored 11 models we track, plus 13 other entries on its leaderboard.
  • Warning Newest score The newest came 2 years ago, in April 2024, for GPT-4 Turbo.
  • Note Frontier models None of the top 10 has a score. The strongest tested is GPT-4 Turbo (#172).
  • Pass Headroom The best is 73.6%, under 85%, so it still separates models.
Leaderboard import 3 pass
  • Pass Importer We read the board on arxiv.org: 11 rows for models we track and 13 other entries at the last read.
  • Pass Last read November 20th, 2026.
  • Pass Read worked The last read worked.
Consistency 1 pass 1 note
  • Note Lab claims No model has both a lab number and an independent run.
  • Pass Known issues No published audit found problems.

How we check