ROME
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
- 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
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:
- GPT-6.2 Astra
- Claude Fable 5.2
-
Gemini 4 Carbon
- GPT-6 Astra
- and 16 more
- 69.4
Gemini 3 Pro
- 61.2
Gemini 2.5 Pro
- 60.2 GPT-5
- 56.4 o3
- 52.7 seed-1-6-vision
- 51.6 o4-mini
- 51.6 GPT-5 mini
- 50.4
Gemini 2.5 Flash
- 50.1 GPT-4.1
- 44.5
Qwen3-VL-235B-A22B-Instruct
- 44.2
Qwen3-VL-235B-A22B
- 42.3 Grok 4
- 41.5
Step 3
- 38.6 Claude Sonnet 4
- 37.8 Qwen-VL-Max
- 36.8 ernie-4.5-vl-424b-a47b
- 34.2 GLM-4.5V
- 33.2 hunyuan-t1-vision-20250619
- 28.8
Qwen2.5-VL-72B
- 28.1
Llama 4 Maverick
- 26.7 QVQ-72B
- 26.3
Mistral Medium 3.1
- 26.0
Mistral Medium 3
- 17.1
Qwen2.5-VL-7B
- 1
Gemini 3 Pro 69.4
- 2
Gemini 2.5 Pro Jun 2025 61.2
- 3 GPT-5 medium 60.2
- 4 o3 medium 56.4
- 5 seed-1-6-vision 52.7
- 6 o4-mini high 51.6
- 7 GPT-5 mini medium 51.6
- 8
Gemini 2.5 Flash Sep 2025, thinking 50.4
- 9 GPT-4.1 50.1
- 10
Qwen3-VL-235B-A22B-Instruct 44.5
- 11
Qwen3-VL-235B-A22B thinking 44.2
- 12 Grok 4 42.3
- 13
Step 3 41.5
- 14 Claude Sonnet 4 thinking 38.6
- 15 Qwen-VL-Max 37.8
- 16 ernie-4.5-vl-424b-a47b 36.8
- 17 GLM-4.5V 34.2
- 18 hunyuan-t1-vision-20250619 33.2
- 19
Qwen2.5-VL-72B 28.8
- 20
Llama 4 Maverick 28.1
- 21 QVQ-72B 26.7
- 22
Mistral Medium 3.1 26.3
- 23
Mistral Medium 3 26.0
- 24
Qwen2.5-VL-7B 17.1
Source: Official ROME leaderboard
- 69.4
Gemini 3 Pro
- 61.2
Gemini 2.5 Pro
- 60.2 GPT-5
- 56.4 o3
- 51.6 o4-mini
- 51.6 GPT-5 mini
- 50.4
Gemini 2.5 Flash
- 50.1 GPT-4.1
- 44.5
Qwen3-VL-235B-A22B-Instruct
- 44.2
Qwen3-VL-235B-A22B
- 42.3 Grok 4
- 41.5
Step 3
- 38.6 Claude Sonnet 4
- 37.8 Qwen-VL-Max
- 34.2 GLM-4.5V
- 28.8
Qwen2.5-VL-72B
- 28.1
Llama 4 Maverick
- 26.3
Mistral Medium 3.1
- 26.0
Mistral Medium 3
- 17.1
Qwen2.5-VL-7B
- 1
Gemini 3 Pro 69.4
- 2
Gemini 2.5 Pro Jun 2025 61.2
- 3 GPT-5 medium 60.2
- 4 o3 medium 56.4
- 5 o4-mini high 51.6
- 6 GPT-5 mini medium 51.6
- 7
Gemini 2.5 Flash Sep 2025, thinking 50.4
- 8 GPT-4.1 50.1
- 9
Qwen3-VL-235B-A22B-Instruct 44.5
- 10
Qwen3-VL-235B-A22B thinking 44.2
- 11 Grok 4 42.3
- 12
Step 3 41.5
- 13 Claude Sonnet 4 thinking 38.6
- 14 Qwen-VL-Max 37.8
- 15 GLM-4.5V 34.2
- 16
Qwen2.5-VL-72B 28.8
- 17
Llama 4 Maverick 28.1
- 18
Mistral Medium 3.1 26.3
- 19
Mistral Medium 3 26.0
- 20
Qwen2.5-VL-7B 17.1
Source: Official ROME leaderboard
| # | Entry | Score ± | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | | 69.4 ±0.9 | Nov 2025 | Official leaderboard | |
| 2 | | 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 | | 50.4 ±1.3 | thinking | Sep 2025 | Official leaderboard |
| 9 | GPT-4.1 | 50.1 ±1.5 | Apr 2025 | Official leaderboard | |
| 10 | | 44.5 | Sep 2025 | Official leaderboard | |
| 11 | | 44.2 | thinking | Sep 2025 | Official leaderboard |
| 12 | Grok 4 | 42.3 ±1.6 | Jul 2025 | Official leaderboard | |
| 13 | | 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 | | 28.8 ±0.2 | Jan 2025 | Official leaderboard | |
| 20 | | 28.1 ±0.9 | Apr 2025 | Official leaderboard | |
| 21 | QVQ-72B | 26.7 ±1.5 | Official leaderboard | ||
| 22 | | 26.3 ±0.5 | Aug 2025 | Official leaderboard | |
| 23 | | 26.0 ±0.8 | May 2025 | Official leaderboard | |
| 24 | | 17.1 ±1.3 | Jan 2025 | Official leaderboard |
| # | Model | Score ± | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | | 69.4 ±0.9 | Nov 2025 | Official leaderboard | |
| 2 | | 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 | | 50.4 ±1.3 | thinking | Sep 2025 | Official leaderboard |
| 8 | GPT-4.1 | 50.1 ±1.5 | Apr 2025 | Official leaderboard | |
| 9 | | 44.5 | Sep 2025 | Official leaderboard | |
| 10 | | 44.2 | thinking | Sep 2025 | Official leaderboard |
| 11 | Grok 4 | 42.3 ±1.6 | Jul 2025 | Official leaderboard | |
| 12 | | 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 | | 28.8 ±0.2 | Jan 2025 | Official leaderboard | |
| 17 | | 28.1 ±0.9 | Apr 2025 | Official leaderboard | |
| 18 | | 26.3 ±0.5 | Aug 2025 | Official leaderboard | |
| 19 | | 26.0 ±0.8 | May 2025 | Official leaderboard | |
| 20 | | 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 GPT-5 ↑2 44.7
- 2
Gemini 3 Pro ↓1 44.7
- 3 o4-mini ↑2 40.9
- 4 GPT-5 mini ↑2 39.4
- 5 o3 ↓1 37.9
- 6
Gemini 2.5 Pro Jun 2025 ↓4 37.9
- 7 Grok 4 ↑4 34.1
- 8 Qwen-VL-Max ↑6 32.7
- 9
Qwen3-VL-235B-A22B ↑1 31.2
- 10
Step 3 ↑2 29.5
Spatial
20 models we track have a score here.
- 1
Gemini 3 Pro 47.1
- 2
Qwen3-VL-235B-A22B-Instruct ↑7 42.9
- 3
Gemini 2.5 Pro Jun 2025 ↓1 42.1
- 4 GPT-5 ↓1 42.1
- 5 o3 ↓1 40.7
- 6 GLM-4.5V ↑9 38.6
- 7 o4-mini ↓2 37.9
- 8 GPT-4.1 37.1
- 9
Qwen3-VL-235B-A22B ↑1 37.1
- 10
Gemini 2.5 Flash Sep 2025 ↓3 36.4
Recognition
20 models we track have a score here.
- 1
Gemini 3 Pro 75.4
- 2
Gemini 2.5 Pro Jun 2025 59.1
- 3 GPT-4.1 ↑5 54.2
- 4 o3 52.3
- 5 GPT-5 ↓2 52.3
- 6 GPT-5 mini 46.2
- 7
Gemini 2.5 Flash Sep 2025 45.8
- 8 o4-mini ↓3 40.2
- 9 Grok 4 ↑2 35.4
- 10
Step 3 ↑2 35.0
Multi
20 models we track have a score here.
- 1
Gemini 3 Pro 70.8
- 2 GPT-5 ↑1 66.7
- 3 o3 ↑1 65.0
- 4 GPT-4.1 ↑4 57.5
- 5 o4-mini 54.2
- 6
Qwen3-VL-235B-A22B-Instruct ↑3 53.3
- 7 Grok 4 ↑4 52.7
- 8
Gemini 2.5 Pro Jun 2025 ↓6 51.7
- 9 GPT-5 mini ↓3 46.7
- 10
Gemini 2.5 Flash Sep 2025 ↓3 46.7
Diagrams
20 models we track have a score here.
- 1
Gemini 3 Pro 77.6
- 2
Gemini 2.5 Pro Jun 2025 65.1
- 3 GPT-5 60.3
- 4
Gemini 2.5 Flash Sep 2025 ↑3 57.0
- 5
Qwen3-VL-235B-A22B ↑5 53.8
- 6 o3 ↓2 52.6
- 7 GPT-5 mini ↓1 52.6
- 8 o4-mini ↓3 49.4
- 9 Claude Sonnet 4 ↑4 48.7
- 10
Qwen3-VL-235B-A22B-Instruct ↓1 48.7
Geo
20 models we track have a score here.
- 1 GPT-5 ↑2 73.0
- 2
Gemini 3 Pro ↓1 71.9
- 3
Gemini 2.5 Pro Jun 2025 ↓1 70.2
- 4 o3 67.5
- 5 GPT-4.1 ↑3 60.8
- 6
Step 3 ↑6 58.2
- 7
Gemini 2.5 Flash Sep 2025 58.0
- 8 o4-mini ↓3 54.0
- 9 GPT-5 mini ↓3 52.8
- 10
Qwen3-VL-235B-A22B-Instruct ↓1 47.7
Academic
20 models we track have a score here.
- 1
Gemini 3 Pro 80.4
- 2
Gemini 2.5 Pro Jun 2025 77.7
- 3 GPT-5 71.6
- 4
Qwen3-VL-235B-A22B ↑6 71.4
- 5 GPT-5 mini ↑1 69.6
- 6 o3 ↓2 68.2
- 7 Claude Sonnet 4 ↑6 68.2
- 8
Qwen3-VL-235B-A22B-Instruct ↑1 67.6
- 9
Gemini 2.5 Flash Sep 2025 ↓2 67.6
- 10 o4-mini ↓5 65.5
Memes
20 models we track have a score here.
- 1
Gemini 3 Pro 86.7
- 2 Grok 4 ↑9 85.0
- 3
Gemini 2.5 Pro Jun 2025 ↓1 82.5
- 4 GPT-5 ↓1 82.5
- 5 o4-mini 78.3
- 6 o3 ↓2 77.5
- 7 GPT-5 mini ↓1 72.5
- 8 GPT-4.1 67.5
- 9
Gemini 2.5 Flash Sep 2025 ↓2 66.7
- 10
Qwen3-VL-235B-A22B 56.7
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.
Source: Official ROME leaderboard
| Model | Effort | Harness | Score | Cost per task | Output tokens |
|---|---|---|---|---|---|
| 69.4 | |||||
| 61.2 | |||||
| GPT-5 | medium | 60.2 | |||
| GPT-5 | high | 59.9 | |||
| GPT-5 | low | 58.1 | |||
| o3 | medium | 56.4 | |||
| o3 | high | 56.2 | |||
| o3 | low | 55.5 | |||
| o4-mini | high | 51.6 | |||
| GPT-5 mini | medium | 51.6 | |||
| thinking | 50.4 | ||||
| GPT-4.1 | 50.1 | ||||
| o4-mini | medium | 49.9 | |||
| o4-mini | low | 47.4 | |||
| 46.1 | |||||
| 44.5 | |||||
| thinking | 44.2 | ||||
| Grok 4 | 42.3 | ||||
| GPT-5 | minimal | 41.7 | |||
| 41.5 | |||||
| Claude Sonnet 4 | thinking | 38.6 | |||
| Qwen-VL-Max | 37.8 | ||||
| Claude Sonnet 4 | 35.6 | ||||
| GLM-4.5V | 34.2 | ||||
| 28.8 | |||||
| 28.1 | |||||
| 26.3 | |||||
| 26.0 | |||||
| 17.1 |
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.
Source: Official ROME leaderboard
| Model | Effort | Harness | Score | Cost per task | Output tokens |
|---|---|---|---|---|---|
| 69.4 | |||||
| 61.2 | |||||
| GPT-5 | medium | 60.2 | |||
| GPT-5 | high | 59.9 | |||
| GPT-5 | low | 58.1 | |||
| o3 | medium | 56.4 | |||
| o3 | high | 56.2 | |||
| o3 | low | 55.5 | |||
| o4-mini | high | 51.6 | |||
| GPT-5 mini | medium | 51.6 | |||
| thinking | 50.4 | ||||
| GPT-4.1 | 50.1 | ||||
| o4-mini | medium | 49.9 | |||
| o4-mini | low | 47.4 | |||
| 46.1 | |||||
| 44.5 | |||||
| thinking | 44.2 | ||||
| Grok 4 | 42.3 | ||||
| GPT-5 | minimal | 41.7 | |||
| 41.5 | |||||
| Claude Sonnet 4 | thinking | 38.6 | |||
| Qwen-VL-Max | 37.8 | ||||
| Claude Sonnet 4 | 35.6 | ||||
| GLM-4.5V | 34.2 | ||||
| 28.8 | |||||
| 28.1 | |||||
| 26.3 | |||||
| 26.0 | |||||
| 17.1 |
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.
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.
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
- 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.