Benchmark Almanac

ROME

Gemini 3 Pro 69.4 leads since November 2025
Not in model cards

Models solve reasoning problems selected to limit contamination.

The publisher reports model results.

Made by
BAAI FlagEval
Released
September 2025
Kind
Language reasoning
Level
University
Scored
Checked automatically
Answers
Short answer
Tasks
All public
Measures
Capability
Language
English
In and out
Text, Images → Text
Interaction
Single turn
Refresh
One fixed set
Run by
Anyone
In views
Vision
Strongest model tested
Gemini 3 Pro #56 on the capability index
Last model added
November 2025
Scores refreshed
November 20th, 2026
24
Entries on its own leaderboard, 20 models we track

Gemini 3 Pro leads with 69.4 among models released by November 2025 Gemini 3 Pro leads with 69.4 among models released by November 2025

Best score per entry from the benchmark's own leaderboard, read on November 20th, 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 November 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 Gemini 3 Pro 69.4
  2. 2 Gemini 2.5 Pro Jun 2025 61.2
  3. 3 GPT-5 medium 60.2
  4. 4 o3 medium 56.4
  5. 5 seed-1-6-vision 52.7
  6. 6 o4-mini high 51.6
  7. 7 GPT-5 mini medium 51.6
  8. 8 Gemini 2.5 Flash Sep 2025, thinking 50.4
  9. 9 GPT-4.1 50.1
  10. 10 Qwen3-VL-235B-A22B-Instruct 44.5
  11. 11 Qwen3-VL-235B-A22B thinking 44.2
  12. 12 Grok 4 42.3
  13. 13 Step 3 41.5
  14. 14 Claude Sonnet 4 thinking 38.6
  15. 15 Qwen-VL-Max 37.8
  16. 16 ernie-4.5-vl-424b-a47b 36.8
  17. 17 GLM-4.5V 34.2
  18. 18 hunyuan-t1-vision-20250619 33.2
  19. 19 Qwen2.5-VL-72B 28.8
  20. 20 Llama 4 Maverick 28.1
  21. 21 QVQ-72B 26.7
  22. 22 Mistral Medium 3.1 26.3
  23. 23 Mistral Medium 3 26.0
  24. 24 Qwen2.5-VL-7B 17.1

Source: Official ROME leaderboard

#EntryScore ±SettingDateSource
1 Gemini 3 Pro 69.4 ±0.9 Nov 2025 Official leaderboard
2 Gemini 2.5 Pro Jun 2025 61.2 ±0.4 Jun 2025 Official leaderboard
3 GPT-5 60.2 ±1.3 medium Aug 2025 Official leaderboard
4 o3 56.4 ±2 medium Apr 2025 Official leaderboard
5 seed-1-6-vision 52.7 ±1.1 Official leaderboard
6 o4-mini 51.6 ±0.8 high Apr 2025 Official leaderboard
7 GPT-5 mini 51.6 ±2.2 medium Aug 2025 Official leaderboard
8 Gemini 2.5 Flash Sep 2025 50.4 ±1.3 thinking Sep 2025 Official leaderboard
9 GPT-4.1 50.1 ±1.5 Apr 2025 Official leaderboard
10 Qwen3-VL-235B-A22B-Instruct 44.5 Sep 2025 Official leaderboard
11 Qwen3-VL-235B-A22B 44.2 thinking Sep 2025 Official leaderboard
12 Grok 4 42.3 ±1.6 Jul 2025 Official leaderboard
13 Step 3 41.5 ±1.5 Jul 2025 Official leaderboard
14 Claude Sonnet 4 38.6 ±1.2 thinking May 2025 Official leaderboard
15 Qwen-VL-Max 37.8 ±1.7 Jan 2024 Official leaderboard
16 ernie-4.5-vl-424b-a47b 36.8 ±2 Official leaderboard
17 GLM-4.5V 34.2 ±1.4 Aug 2025 Official leaderboard
18 hunyuan-t1-vision-20250619 33.2 ±1.6 Official leaderboard
19 Qwen2.5-VL-72B 28.8 ±0.2 Jan 2025 Official leaderboard
20 Llama 4 Maverick 28.1 ±0.9 Apr 2025 Official leaderboard
21 QVQ-72B 26.7 ±1.5 Official leaderboard
22 Mistral Medium 3.1 26.3 ±0.5 Aug 2025 Official leaderboard
23 Mistral Medium 3 26.0 ±0.8 May 2025 Official leaderboard
24 Qwen2.5-VL-7B 17.1 ±1.3 Jan 2025 Official leaderboard
#ModelScore ±SettingDateSource
1 Gemini 3 Pro 69.4 ±0.9 Nov 2025 Official leaderboard
2 Gemini 2.5 Pro Jun 2025 61.2 ±0.4 Jun 2025 Official leaderboard
3 GPT-5 60.2 ±1.3 medium Aug 2025 Official leaderboard
4 o3 56.4 ±2 medium Apr 2025 Official leaderboard
5 o4-mini 51.6 ±0.8 high Apr 2025 Official leaderboard
6 GPT-5 mini 51.6 ±2.2 medium Aug 2025 Official leaderboard
7 Gemini 2.5 Flash Sep 2025 50.4 ±1.3 thinking Sep 2025 Official leaderboard
8 GPT-4.1 50.1 ±1.5 Apr 2025 Official leaderboard
9 Qwen3-VL-235B-A22B-Instruct 44.5 Sep 2025 Official leaderboard
10 Qwen3-VL-235B-A22B 44.2 thinking Sep 2025 Official leaderboard
11 Grok 4 42.3 ±1.6 Jul 2025 Official leaderboard
12 Step 3 41.5 ±1.5 Jul 2025 Official leaderboard
13 Claude Sonnet 4 38.6 ±1.2 thinking May 2025 Official leaderboard
14 Qwen-VL-Max 37.8 ±1.7 Jan 2024 Official leaderboard
15 GLM-4.5V 34.2 ±1.4 Aug 2025 Official leaderboard
16 Qwen2.5-VL-72B 28.8 ±0.2 Jan 2025 Official leaderboard
17 Llama 4 Maverick 28.1 ±0.9 Apr 2025 Official leaderboard
18 Mistral Medium 3.1 26.3 ±0.5 Aug 2025 Official leaderboard
19 Mistral Medium 3 26.0 ±0.8 May 2025 Official leaderboard
20 Qwen2.5-VL-7B 17.1 ±1.3 Jan 2025 Official leaderboard

Rankings by part

We read 8 parts from the benchmark’s own leaderboard, each ranking models on one part of the test, such as Puzzles Games, Spatial and Recognition. Gemini 3 Pro leads 6 of them, and GPT-5 leads the other 2. Arrows show how far a model moved from its overall rank among the models we track.

Puzzles Games

20 models we track have a score here.

  1. 1 GPT-5 ↑2 44.7
  2. 2 Gemini 3 Pro ↓1 44.7
  3. 3 o4-mini ↑2 40.9
  4. 4 GPT-5 mini ↑2 39.4
  5. 5 o3 ↓1 37.9
  6. 6 Gemini 2.5 Pro Jun 2025 ↓4 37.9
  7. 7 Grok 4 ↑4 34.1
  8. 8 Qwen-VL-Max ↑6 32.7
  9. 9 Qwen3-VL-235B-A22B ↑1 31.2
  10. 10 Step 3 ↑2 29.5

More effort adds a median 4 score

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
GPT-5
minimal low medium
60.2
o3
low
56.4
o4-mini
low medium
51.6
Gemini 2.5 Flash
thinking
50.4
Claude Sonnet 4
thinking
38.6

Source: Official ROME leaderboard

More effort adds a median 4 score

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
GPT-5
minimal low medium
60.2
o3
low
56.4
o4-mini
low medium
51.6
Gemini 2.5 Flash
thinking
50.4
Claude Sonnet 4
thinking
38.6

Source: Official ROME leaderboard

2 models have held number one on the official leaderboard since its release in September 2025

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

Google DeepMind
Own leaderboard Not tested since Nov 2025

Source: Official ROME 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
0.00 20.0 40.0 60.0 80.0 2024 2025 2026 0.00 20.0 40.0 60.0 80.0 2024 2025 2026

Source: Official ROME 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 Knowledge

All 332 in Knowledge
  • MMLU-Pro In use Harder 10-option version of MMLU with reasoning-focused questions across 14 subjects. Intern-S2-397B 89.8%
  • MMLU In use 57-subject multiple-choice exam questions from elementary to professional level. Scored by accuracy. Seed1.8 92.3%
  • SuperGPQA In use 26,529 graduate-level multiple-choice questions across 285 disciplines. Accuracy. Qwen 3.6 Max (Preview) 73.9%
  • WideSearch In use 200 bilingual broad info-seeking tasks: agent collects many atomic facts into a table. Scored by success rate and row/item F1. Tencent Hy4 preview 83.9%
  • BBEH micro average In use Models solve harder replacements for each of the 23 BIG-Bench Hard tasks, 4,520 examples in total. The score is the share of all examples answered correctly. Gemma 4 31B IT 74.4%
  • HLE-Diamond In use Models answer 1,000 cleaned expert questions from Humanity's Last Exam, 500 on reasoning and 500 on knowledge, closed book. Graded on exact answers. Claude Fable 5.2 77.9%

Data audit

8 of 9 checks pass, 1 warning. Look first at the Newest score check. The newest came 12 months ago, in November 2025, for Gemini 3 Pro. 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 (24 entries).
  • Pass Models scored 20 models we track, plus 4 other entries on its leaderboard.
  • Warning Newest score The newest came 12 months ago, in November 2025, for Gemini 3 Pro.
  • Note Frontier models None of the top 10 has a score. The strongest tested is Gemini 3 Pro (#56).
  • Note Headroom Scored as a plain number, with no fixed ceiling.
Leaderboard import 3 pass
  • Pass Importer We read the board on flageval-baai.github.io: 29 rows for models we track and 4 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