Benchmark Almanac

FlagEval Arena

o4-mini 1,134 leads since April 2025
Not in model cards

Models answer the same prompt anonymously. Users compare their responses, and the leaderboard estimates a rating from the votes.

The official ratings feed lists 36 models.

Kind
Overall chat quality
Level
Everyday
Scored
Ranked by votes
Answers
Open-ended
Tasks
All public
Measures
Preference
Language
Chinese
In and out
Text → Text
Interaction
Single turn
Refresh
Live
Run by
A host
In views
Kept fresh
Strongest model tested
o3 #84 on the capability index
Last model added
September 2025
Scores refreshed
November 21st, 2026
31
Entries on its own leaderboard, 28 models we track

o4-mini leads with 1,134 among models released by September 2025 o4-mini leads with 1,134 among models released by September 2025

Best score per entry from the benchmark's own leaderboard, read on November 21st, 2026. Where a model ran with several settings or harnesses, we keep its best. Every run is in the next chart. 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 September 2025, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

Bars start at 800.

  1. 1 o4-mini 1,134
  2. 2 Magistral Medium 1.0 1,118
  3. 3 GPT-4.1 mini 1,107
  4. 4 o3 1,106
  5. 5 Gemini 2.5 Pro Jun 2025 1,092
  6. 6 Qwen3-235B-A22B 1,063
  7. 7 Qwen3-235B-A22B-Instruct Jul 2025 1,049
  8. 8 Claude 3.7 Sonnet 1,043
  9. 9 Doubao-Seed-1.6 1,042
  10. 10 Claude Sonnet 4 1,035
  11. 11 ERNIE-4.5-300B-A47B 1,032
  12. 12 Hunyuan-TurboS 1,029
  13. 13 GLM-4.5 1,027
  14. 14 DeepSeek-V3 Mar 2025 1,019
  15. 15 Mistral Medium 3.1 1,011
  16. 16 DeepSeek-V3.1 1,008
  17. 17 Qwen3-235B-A22B-Thinking Jul 2025 1,006
  18. 18 GPT-4.1 1,002
  19. 19 DeepSeek-R1 May 2025 993
  20. 20 Kimi K2 Sep 2025 986
  21. 21 Gemini 2.5 Flash Sep 2025 983
  22. 22 Qwen2.5-Max 981
  23. 23 Grok 4 974
  24. 24 GPT-4o (2024-11-20) 973
  25. 25 Doubao-1.5-pro 961
  26. 26 Mistral Medium 3 956
  27. 27 Yi-Lightning 953
  28. 28 GLM-4-plus 945
  29. 29 Qwen2.5-72B 927
  30. 30 MiniMax-M1-80k 909
  31. 31 GLM-Zero-Preview 876

Source: Official FlagEval Arena leaderboard

#EntryScore ±SettingDateSource
1 o4-mini 1,134 1127 to 1142.6 Apr 2025 Official leaderboard
2 Magistral Medium 1.0 1,118 1112.2 to 1124 Jun 2025 Official leaderboard
3 GPT-4.1 mini 1,107 1100.6 to 1111.8 Apr 2025 Official leaderboard
4 o3 1,106 1097.7 to 1114 Apr 2025 Official leaderboard
5 Gemini 2.5 Pro Jun 2025 1,092 1080.8 to 1100.7 Jun 2025 Official leaderboard
6 Qwen3-235B-A22B 1,063 1058.8 to 1067.7 Apr 2025 Official leaderboard
7 Qwen3-235B-A22B-Instruct Jul 2025 1,049 1038.3 to 1059.5 Jul 2025 Official leaderboard
8 Claude 3.7 Sonnet 1,043 1039.7 to 1047.6 Feb 2025 Official leaderboard
9 Doubao-Seed-1.6 1,042 1033.5 to 1049.4 Jun 2025 Official leaderboard
10 Claude Sonnet 4 1,035 ±7.1 May 2025 Official leaderboard
11 ERNIE-4.5-300B-A47B 1,032 1023.7 to 1040.5 Jun 2025 Official leaderboard
12 Hunyuan-TurboS 1,029 1022.7 to 1037.4 Feb 2025 Official leaderboard
13 GLM-4.5 1,027 1015.9 to 1036.8 Aug 2025 Official leaderboard
14 DeepSeek-V3 Mar 2025 1,019 1012 to 1027.5 Mar 2025 Official leaderboard
15 Mistral Medium 3.1 1,011 ±12.2 Aug 2025 Official leaderboard
16 DeepSeek-V3.1 1,008 993.2 to 1019 Aug 2025 Official leaderboard
17 Qwen3-235B-A22B-Thinking Jul 2025 1,006 996.7 to 1016.7 Jul 2025 Official leaderboard
18 GPT-4.1 1,002 994.3 to 1009.5 Apr 2025 Official leaderboard
19 DeepSeek-R1 May 2025 993 981.6 to 1006.2 May 2025 Official leaderboard
20 Kimi K2 Sep 2025 986 ±8.8 Sep 2025 Official leaderboard
21 Gemini 2.5 Flash Sep 2025 983 975.9 to 991.6 Sep 2025 Official leaderboard
22 Qwen2.5-Max 981 974.9 to 987.3 Jan 2025 Official leaderboard
23 Grok 4 974 968.3 to 978.2 Jul 2025 Official leaderboard
24 GPT-4o (2024-11-20) 973 968.1 to 977.9 Official leaderboard
25 Doubao-1.5-pro 961 ±6.4 Jan 2025 Official leaderboard
26 Mistral Medium 3 956 945.3 to 964.7 May 2025 Official leaderboard
27 Yi-Lightning 953 ±4.1 Dec 2024 Official leaderboard
28 GLM-4-plus 945 940.7 to 950.1 Official leaderboard
29 Qwen2.5-72B 927 ±7 Sep 2024 Official leaderboard
30 MiniMax-M1-80k 909 904.7 to 914.4 Jun 2025 Official leaderboard
31 GLM-Zero-Preview 876 870.9 to 880.3 Official leaderboard
#ModelScore ±SettingDateSource
1 o4-mini 1,134 1127 to 1142.6 Apr 2025 Official leaderboard
2 Magistral Medium 1.0 1,118 1112.2 to 1124 Jun 2025 Official leaderboard
3 GPT-4.1 mini 1,107 1100.6 to 1111.8 Apr 2025 Official leaderboard
4 o3 1,106 1097.7 to 1114 Apr 2025 Official leaderboard
5 Gemini 2.5 Pro Jun 2025 1,092 1080.8 to 1100.7 Jun 2025 Official leaderboard
6 Qwen3-235B-A22B 1,063 1058.8 to 1067.7 Apr 2025 Official leaderboard
7 Qwen3-235B-A22B-Instruct Jul 2025 1,049 1038.3 to 1059.5 Jul 2025 Official leaderboard
8 Claude 3.7 Sonnet 1,043 1039.7 to 1047.6 Feb 2025 Official leaderboard
9 Doubao-Seed-1.6 1,042 1033.5 to 1049.4 Jun 2025 Official leaderboard
10 Claude Sonnet 4 1,035 ±7.1 May 2025 Official leaderboard
11 ERNIE-4.5-300B-A47B 1,032 1023.7 to 1040.5 Jun 2025 Official leaderboard
12 Hunyuan-TurboS 1,029 1022.7 to 1037.4 Feb 2025 Official leaderboard
13 GLM-4.5 1,027 1015.9 to 1036.8 Aug 2025 Official leaderboard
14 DeepSeek-V3 Mar 2025 1,019 1012 to 1027.5 Mar 2025 Official leaderboard
15 Mistral Medium 3.1 1,011 ±12.2 Aug 2025 Official leaderboard
16 DeepSeek-V3.1 1,008 993.2 to 1019 Aug 2025 Official leaderboard
17 Qwen3-235B-A22B-Thinking Jul 2025 1,006 996.7 to 1016.7 Jul 2025 Official leaderboard
18 GPT-4.1 1,002 994.3 to 1009.5 Apr 2025 Official leaderboard
19 DeepSeek-R1 May 2025 993 981.6 to 1006.2 May 2025 Official leaderboard
20 Kimi K2 Sep 2025 986 ±8.8 Sep 2025 Official leaderboard
21 Gemini 2.5 Flash Sep 2025 983 975.9 to 991.6 Sep 2025 Official leaderboard
22 Qwen2.5-Max 981 974.9 to 987.3 Jan 2025 Official leaderboard
23 Grok 4 974 968.3 to 978.2 Jul 2025 Official leaderboard
24 Doubao-1.5-pro 961 ±6.4 Jan 2025 Official leaderboard
25 Mistral Medium 3 956 945.3 to 964.7 May 2025 Official leaderboard
26 Yi-Lightning 953 ±4.1 Dec 2024 Official leaderboard
27 Qwen2.5-72B 927 ±7 Sep 2024 Official leaderboard
28 MiniMax-M1-80k 909 904.7 to 914.4 Jun 2025 Official leaderboard

Scores at each effort level

Models the leaderboard ran at two or more effort levels. We don’t have enough cost data here to say what the extra effort costs.

Lowest effort to highest, light to dark
o4-mini
high
1,134
Magistral Medium 1.0
thinking
1,118
Claude 3.7 Sonnet
thinking
1,043
Doubao-Seed-1.6
thinking
1,042

Source: Official FlagEval Arena leaderboard

Scores at each effort level

Models the leaderboard ran at two or more effort levels. We don’t have enough cost data here to say what the extra effort costs.

Lowest effort to highest, light to dark
o4-mini
high
1,134
Magistral Medium 1.0
thinking
1,118
Claude 3.7 Sonnet
thinking
1,043
Doubao-Seed-1.6
thinking
1,042

Source: Official FlagEval Arena leaderboard

7 models have held number one on the official leaderboard since 2023

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

Qwen 01.AI ByteDance Seed Anthropic OpenAI
Own leaderboard Not tested since Sep 2025
o4-mini 1,134

Source: Official FlagEval Arena leaderboard

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
On the benchmark’s own leaderboard
800 900 1,000 1,100 1,200 2025 800 900 1,000 1,100 1,200 2025

Source: Official FlagEval Arena 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 Instructions and chat

All 144 in Instructions and chat
  • IFEval In use Models answer about 500 prompts containing verifiable instructions such as word limits or required keywords. Code checks each instruction for strict and loose accuracy. Gemma 4 31B IT 98.9%
  • IFBench In use 58 new out-of-domain verifiable output constraints. Scored by programmatic constraint checks (strict or loose). Flint 89.2%
  • MultiChallenge In use Multi-turn conversations testing instruction retention, inference memory, versioned editing and self-coherence. Graded by LLM judge with rubrics. GPT-6.1 Sol 82%
  • Multi-IF In use Models follow verifiable instructions over three-turn conversations in eight languages, extending IFEval. Scored by instruction accuracy at each turn. LFM3-1.7B 89.1%
  • AdvancedIF In use Models respond to over 1,600 prompts with complex, multi-turn, and system-level instructions. A judge model checks each response against expert-written rubrics. MAI-Thinking-1 85%
  • MM-IFEval In use Multimodal models follow image-grounded instructions with exact output constraints. Checks mix rule-based verification and a judge model. GPT-4o 64.6%

Data audit

8 of 9 checks pass, 1 warning. Look first at the Newest score check. The newest came 14 months ago, in September 2025, for Gemini 2.5 Flash. Every benchmark

Catalog record 2 pass
  • Pass Task access All tasks are public.
  • Pass Confidence Checked against a primary source.
Scores 1 warning 2 pass 2 notes
  • Pass Independent runs Scores from the official leaderboard (31 entries).
  • Pass Models scored 28 models we track, plus 3 other entries on its leaderboard.
  • Warning Newest score The newest came 14 months ago, in September 2025, for Gemini 2.5 Flash.
  • Note Frontier models None of the top 10 has a score. The strongest tested is o3 (#84).
  • Note Headroom Scored as a rating, with no fixed ceiling.
Leaderboard import 3 pass
  • Pass Importer We read the board on flageval.baai.ac.cn: 33 rows for models we track and 3 other entries at the last read.
  • Pass Last read November 21st, 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