Benchmark Almanac

TMMLU+

GPT-5 88.6% leads since August 2025
Not in model cards
Version 1.0

Models answer traditional Chinese questions across 66 subjects. The evaluation includes elementary material and professional examinations.

Made by
iKala
Released
December 2023
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
Run by
Anyone
Strongest model tested
GPT-5 #71 on the capability index
Last model added
August 2025
Scores refreshed
November 20th, 2026
47
Entries in its paper, 19 models we track

GPT-5 leads with 88.6% among models released by August 2025 GPT-5 leads with 88.6% among models released by August 2025

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 August 2025, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

  1. 1 GPT-5 v1.0, zero-shot direct answering 88.6%
  2. 2 gpt-oss-120b v1.0, zero-shot direct answering 69.1%
  3. 3 Gemini-1.5-pro (snapshot unspecified) v1.0, zero-shot direct answering 64.7%
  4. 4 Qwen/Qwen-72B v1.0, zero-shot direct answering 64.3%
  5. 5 GPT-4 Jun 2023, v1.0, zero-shot direct answering 60.3%
  6. 6 Qwen-max (snapshot unspecified) v1.0, zero-shot direct answering 60.2%
  7. 7 Qwen/Qwen-72B-Chat v1.0, zero-shot direct answering 58.5%
  8. 8 gpt-oss-20b v1.0, zero-shot direct answering 56.6%
  9. 9 Qwen/Qwen-14B [base checkpoint] v1.0, zero-shot direct answering 50.5%
  10. 10 Gemini 1.0 Pro v1.0, zero-shot direct answering 49.9%
  11. 11 01-ai/Yi-34B-Chat v1.0, zero-shot direct answering 49.6%
  12. 12 Gemini-1.5-flash (snapshot unspecified) v1.0, zero-shot direct answering 49.1%
  13. 13 Reka Flash v1.0, zero-shot direct answering 47.1%
  14. 14 Qwen-14B v1.0, zero-shot direct answering 46.8%
  15. 15 Qwen1.5-14B-Chat v1.0, zero-shot direct answering 45.3%
  16. 16 Yi-6B-Chat v1.0, zero-shot direct answering 44.6%
  17. 17 Claude 1.3 v1.0, zero-shot direct answering 44.6%
  18. 18 MediaTek-Research/Breeze-7B-Instruct-v0_1 v1.0, zero-shot direct answering 42.7%
  19. 19 GPT-3.5 Turbo Jun 2023, v1.0, zero-shot direct answering 41.8%
  20. 20 CausalLM/14B v1.0, zero-shot direct answering 41.5%
  21. 21 Skywork/Skywork-13B-base v1.0, zero-shot direct answering 41.3%
  22. 22 Claude 3 Opus v1.0, zero-shot direct answering 41.1%
  23. 23 Qwen-7B v1.0, zero-shot direct answering 40%
  24. 24 meta-llama/Llama-3-70b-chat-hf v1.0, zero-shot direct answering 39.6%
  25. 25 Qwen/Qwen-7B-Chat v1.0, zero-shot direct answering 39.5%
  26. 26 vivo-ai/BlueLM-7B-Base v1.0, zero-shot direct answering 37.9%
  27. 27 Baichuan2-13B-Chat v1.0, zero-shot direct answering 37.6%
  28. 28 Qwen-1_8B v1.0, zero-shot direct answering 36.3%
  29. 29 Claude 2 v1.0, zero-shot direct answering 36.2%
  30. 30 ChatGLM3-6B v1.0, zero-shot direct answering 35.4%
  31. 31 deepseek-ai/deepseek-llm-7b-chat v1.0, zero-shot direct answering 35.2%
  32. 32 Mistral 7B v0.1 v1.0, zero-shot direct answering 35.1%
  33. 33 Azure99/blossom-v3_1-mistral-7b v1.0, zero-shot direct answering 34.1%
  34. 34 google/gemma-7b-it v1.0, zero-shot direct answering 33.8%
  35. 35 Reka Edge [TMMLU+ v1.0 release] v1.0, zero-shot direct answering 33.4%
  36. 36 microsoft/Orca-2-13b v1.0, zero-shot direct answering 32.4%
  37. 37 Qwen/Qwen-1_8B-Chat v1.0, zero-shot direct answering 31.7%
  38. 38 meta-llama/Llama-3-8b-chat-hf v1.0, zero-shot direct answering 31.6%
  39. 39 TigerResearch/tigerbot-13b-chat-v3 v1.0, zero-shot direct answering 26.8%
  40. 40 hongyin/mistral-7b-80k v1.0, zero-shot direct answering 23.8%
  41. 41 DeepSeek LLM 67B v1.0, zero-shot direct answering 22.1%
  42. 42 yentinglin/Taiwan-LLM-13B-v2.0-chat v1.0, zero-shot direct answering 21.4%
  43. 43 GeneZC/MiniChat-3B v1.0, zero-shot direct answering 21%
  44. 44 LinkSoul/Chinese-Llama-2-7b v1.0, zero-shot direct answering 16%
  45. 45 yentinglin/Taiwan-LLM-7B-v2.1-chat v1.0, zero-shot direct answering 15.6%
  46. 46 Claude-instant-1 v1.0, zero-shot direct answering 14.6%
  47. 47 FlagAlpha/Atom-7B v1.0, zero-shot direct answering 9.68%

Source: The TMMLU+ paper

#EntryScoreSettingDateSource
1 GPT-5 88.6% v1.0, zero-shot direct answering Aug 2025 Official leaderboard
2 gpt-oss-120b 69.1% v1.0, zero-shot direct answering Aug 2025 Official leaderboard
3 Gemini-1.5-pro (snapshot unspecified) 64.7% v1.0, zero-shot direct answering Official leaderboard
4 Qwen/Qwen-72B 64.3% v1.0, zero-shot direct answering Official leaderboard
5 GPT-4 Jun 2023 60.3% v1.0, zero-shot direct answering Jun 2023 Official leaderboard
6 Qwen-max (snapshot unspecified) 60.2% v1.0, zero-shot direct answering Official leaderboard
7 Qwen/Qwen-72B-Chat 58.5% v1.0, zero-shot direct answering Official leaderboard
8 gpt-oss-20b 56.6% v1.0, zero-shot direct answering Aug 2025 Official leaderboard
9 Qwen/Qwen-14B [base checkpoint] 50.5% v1.0, zero-shot direct answering Official leaderboard
10 Gemini 1.0 Pro 49.9% v1.0, zero-shot direct answering Dec 2023 Official leaderboard
11 01-ai/Yi-34B-Chat 49.6% v1.0, zero-shot direct answering Official leaderboard
12 Gemini-1.5-flash (snapshot unspecified) 49.1% v1.0, zero-shot direct answering Official leaderboard
13 Reka Flash 47.1% v1.0, zero-shot direct answering Feb 2024 Official leaderboard
14 Qwen-14B 46.8% v1.0, zero-shot direct answering Sep 2023 Official leaderboard
15 Qwen1.5-14B-Chat 45.3% v1.0, zero-shot direct answering Feb 2024 Official leaderboard
16 Yi-6B-Chat 44.6% v1.0, zero-shot direct answering Nov 2023 Official leaderboard
17 Claude 1.3 44.6% v1.0, zero-shot direct answering Apr 2023 Official leaderboard
18 MediaTek-Research/Breeze-7B-Instruct-v0_1 42.7% v1.0, zero-shot direct answering Official leaderboard
19 GPT-3.5 Turbo Jun 2023 41.8% v1.0, zero-shot direct answering Jun 2023 Official leaderboard
20 CausalLM/14B 41.5% v1.0, zero-shot direct answering Official leaderboard
21 Skywork/Skywork-13B-base 41.3% v1.0, zero-shot direct answering Official leaderboard
22 Claude 3 Opus 41.1% v1.0, zero-shot direct answering Feb 2024 Official leaderboard
23 Qwen-7B 40% v1.0, zero-shot direct answering Sep 2023 Official leaderboard
24 meta-llama/Llama-3-70b-chat-hf 39.6% v1.0, zero-shot direct answering Official leaderboard
25 Qwen/Qwen-7B-Chat 39.5% v1.0, zero-shot direct answering Official leaderboard
26 vivo-ai/BlueLM-7B-Base 37.9% v1.0, zero-shot direct answering Official leaderboard
27 Baichuan2-13B-Chat 37.6% v1.0, zero-shot direct answering Sep 2023 Official leaderboard
28 Qwen-1_8B 36.3% v1.0, zero-shot direct answering Nov 2023 Official leaderboard
29 Claude 2 36.2% v1.0, zero-shot direct answering Jul 2023 Official leaderboard
30 ChatGLM3-6B 35.4% v1.0, zero-shot direct answering Oct 2023 Official leaderboard
31 deepseek-ai/deepseek-llm-7b-chat 35.2% v1.0, zero-shot direct answering Official leaderboard
32 Mistral 7B v0.1 35.1% v1.0, zero-shot direct answering Sep 2023 Official leaderboard
33 Azure99/blossom-v3_1-mistral-7b 34.1% v1.0, zero-shot direct answering Official leaderboard
34 google/gemma-7b-it 33.8% v1.0, zero-shot direct answering Official leaderboard
35 Reka Edge [TMMLU+ v1.0 release] 33.4% v1.0, zero-shot direct answering Official leaderboard
36 microsoft/Orca-2-13b 32.4% v1.0, zero-shot direct answering Official leaderboard
37 Qwen/Qwen-1_8B-Chat 31.7% v1.0, zero-shot direct answering Official leaderboard
38 meta-llama/Llama-3-8b-chat-hf 31.6% v1.0, zero-shot direct answering Official leaderboard
39 TigerResearch/tigerbot-13b-chat-v3 26.8% v1.0, zero-shot direct answering Official leaderboard
40 hongyin/mistral-7b-80k 23.8% v1.0, zero-shot direct answering Official leaderboard
41 DeepSeek LLM 67B 22.1% v1.0, zero-shot direct answering Nov 2023 Official leaderboard
42 yentinglin/Taiwan-LLM-13B-v2.0-chat 21.4% v1.0, zero-shot direct answering Official leaderboard
43 GeneZC/MiniChat-3B 21% v1.0, zero-shot direct answering Official leaderboard
44 LinkSoul/Chinese-Llama-2-7b 16% v1.0, zero-shot direct answering Official leaderboard
45 yentinglin/Taiwan-LLM-7B-v2.1-chat 15.6% v1.0, zero-shot direct answering Official leaderboard
46 Claude-instant-1 14.6% v1.0, zero-shot direct answering Official leaderboard
47 FlagAlpha/Atom-7B 9.68% v1.0, zero-shot direct answering Official leaderboard
#ModelScoreSettingDateSource
1 GPT-5 88.6% v1.0, zero-shot direct answering Aug 2025 Official leaderboard
2 gpt-oss-120b 69.1% v1.0, zero-shot direct answering Aug 2025 Official leaderboard
3 GPT-4 Jun 2023 60.3% v1.0, zero-shot direct answering Jun 2023 Official leaderboard
4 gpt-oss-20b 56.6% v1.0, zero-shot direct answering Aug 2025 Official leaderboard
5 Gemini 1.0 Pro 49.9% v1.0, zero-shot direct answering Dec 2023 Official leaderboard
6 Reka Flash 47.1% v1.0, zero-shot direct answering Feb 2024 Official leaderboard
7 Qwen-14B 46.8% v1.0, zero-shot direct answering Sep 2023 Official leaderboard
8 Qwen1.5-14B-Chat 45.3% v1.0, zero-shot direct answering Feb 2024 Official leaderboard
9 Yi-6B-Chat 44.6% v1.0, zero-shot direct answering Nov 2023 Official leaderboard
10 Claude 1.3 44.6% v1.0, zero-shot direct answering Apr 2023 Official leaderboard
11 GPT-3.5 Turbo Jun 2023 41.8% v1.0, zero-shot direct answering Jun 2023 Official leaderboard
12 Claude 3 Opus 41.1% v1.0, zero-shot direct answering Feb 2024 Official leaderboard
13 Qwen-7B 40% v1.0, zero-shot direct answering Sep 2023 Official leaderboard
14 Baichuan2-13B-Chat 37.6% v1.0, zero-shot direct answering Sep 2023 Official leaderboard
15 Qwen-1_8B 36.3% v1.0, zero-shot direct answering Nov 2023 Official leaderboard
16 Claude 2 36.2% v1.0, zero-shot direct answering Jul 2023 Official leaderboard
17 ChatGLM3-6B 35.4% v1.0, zero-shot direct answering Oct 2023 Official leaderboard
18 Mistral 7B v0.1 35.1% v1.0, zero-shot direct answering Sep 2023 Official leaderboard
19 DeepSeek LLM 67B 22.1% v1.0, zero-shot direct answering Nov 2023 Official leaderboard

Rankings by part

We read 4 parts from the benchmark’s paper, each ranking models on one part of the test, such as STEM, Social Science and Humanities. GPT-5 leads every one of them. Arrows show how far a model moved from its overall rank among the models we track.

STEM

19 models we track have a score here.

  1. 1 GPT-5 92.8%
  2. 2 gpt-oss-120b 82.1%
  3. 3 gpt-oss-20b ↑1 73.9%
  4. 4 GPT-4 Jun 2023 ↓1 60.4%
  5. 5 Gemini 1.0 Pro 45.4%
  6. 6 Reka Flash 45.3%
  7. 7 Qwen-14B 43.9%
  8. 8 Claude 3 Opus ↑4 43%
  9. 9 Claude 1.3 ↑1 42.6%
  10. 10 GPT-3.5 Turbo Jun 2023 ↑1 41.6%

3 models have held number one in the benchmark’s paper since its release in December 2023

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

OpenAI
Benchmark paper Passed 85% in Aug 2025, by GPT-5
GPT-5 88.6%

Source: The TMMLU+ paper

GPT-5 on the official leaderboard passed 85% in August 2025

That was 20 months after its release. By the same age, 36% of the benchmarks from 2023 had passed 85%. 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% 2023 2024 2025 Near the ceiling (85%) GPT-5 gpt-oss-120b GPT-4 Claude 1.3 0% 20% 40% 60% 80% 100% 2023 2024 2025 Near the ceiling (85%) GPT-5 gpt-oss-120b

Source: Official TMMLU+ 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.

Versions

We switch the top row to each new version when it comes out. Open it for one row per version. An old version stays in play while people still test it, so whoever first takes it past 85% gets credit for closing it out.

OpenAI Anthropic
TMMLU+ 1.1 Passed 85% in Sep 2026, by Claude Opus 5
TMMLU+ Passed 85% in Aug 2025, by GPT-5
GPT-5 88.6%
TMMLU+ 1.1 Passed 85% in Sep 2026, by Claude Opus 5

Source: Official leaderboards

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

All 8 checks pass. Every benchmark

Catalog record 2 pass
  • Pass Task access All tasks are public.
  • Pass Confidence Checked against a primary source.
Scores 2 pass 3 notes
  • Pass Independent runs Scores from the benchmark’s paper (47 entries).
  • Pass Models scored 19 models we track, plus 28 other entries on its leaderboard.
  • Note Newest score The newest came 15 months ago, in August 2025, for GPT-5. That’s expected for a test replaced by TMMLU+ 1.1.
  • Note Frontier models None of the top 10 has a score, which is fine for a retired test. The strongest tested is GPT-5 (#71).
  • Note Headroom Saturated. The best is 88.6%, past 85% since August 2025.
Leaderboard import 3 pass
  • Pass Importer We read the board on huggingface.co: 19 rows for models we track and 28 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