MIRAGE-MMST
Models answer growers’ questions using photographs and written descriptions. They identify the cause of a problem and recommend how to manage it.
Three model judges grade identification and management. The standard leaderboard averages those two scores on a 100-point scale.
- Made by
- University of Illinois Urbana-Champaign and Amazon
- Released
- June 2025
- Kind
- General visual questions
- Level
- University
- Scored
- Graded by a model
- Answers
- Open-ended
- Tasks
- All public
- Measures
- Capability
- Language
- English
- Interaction
- Single turn
- Items from
- Real work and use
- Refresh
- One fixed set
- Run by
- Anyone
- In views
- Vision
- Strongest model tested
- Claude 3.7 Sonnet #119 on the capability index
- Last model added
- April 2025
- Scores refreshed
- November 20th, 2026
GPT-4.1 leads with 68.0 among models released by April 2025
Best score per model from the benchmark's own leaderboard, read on November 20th, 2026. Where a model ran with several settings or harnesses, we keep its best.
The newest model with a score here came out in April 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
- 68.0 GPT-4.1
- 60.7 GPT-4.1 mini
- 58.8 GPT-4o
- 58.4 Claude 3.7 Sonnet
- 56.5 Claude 3.5 Sonnet
- 54.5
Qwen2.5-VL-72B
- 54.1
Qwen2.5-VL-32B
- 50.5
Gemma 3 27B
- 50.2
InternVL3-78B
- 50.0 GPT-4o mini
- 48.0
InternVL3-38B
- 47.5
Llama 4 Scout
- 46.8
Gemma 3 12B
- 44.9
Qwen2.5-VL-7B
- 44.8 Claude 3 Haiku
- 44.5
InternVL3-14B
- 41.2
InternVL3-8B
- 40.5
Gemma 3 4B
- 38.0
Qwen2.5-VL-3B
- 37.8
LLaVA-OneVision 7B
- 36.3
InternVL3-2B
- 35.4 llava-v1.6-mistral-7b
- 1 GPT-4.1 68.0
- 2 GPT-4.1 mini 60.7
- 3 GPT-4o Mar 2025 58.8
- 4 Claude 3.7 Sonnet 58.4
- 5 Claude 3.5 Sonnet 56.5
- 6
Qwen2.5-VL-72B 54.5
- 7
Qwen2.5-VL-32B 54.1
- 8
Gemma 3 27B 50.5
- 9
InternVL3-78B 50.2
- 10 GPT-4o mini 50.0
- 11
InternVL3-38B 48.0
- 12
Llama 4 Scout 47.5
- 13
Gemma 3 12B 46.8
- 14
Qwen2.5-VL-7B 44.9
- 15 Claude 3 Haiku 44.8
- 16
InternVL3-14B 44.5
- 17
InternVL3-8B 41.2
- 18
Gemma 3 4B 40.5
- 19
Qwen2.5-VL-3B 38.0
- 20
LLaVA-OneVision 7B 37.8
- 21
InternVL3-2B 36.3
- 22 llava-v1.6-mistral-7b 35.4
Source: Official MIRAGE-MMST leaderboard
| # | Model | Score | Setting | Released | Run on | Source |
|---|---|---|---|---|---|---|
| 1 | GPT-4.1 | 68.0 | Apr 2025 | Jun 2025 | Official leaderboard | |
| 2 | GPT-4.1 mini | 60.7 | Apr 2025 | Jun 2025 | Official leaderboard | |
| 3 | GPT-4o Mar 2025 | 58.8 | Mar 2025 | Jun 2025 | Official leaderboard | |
| 4 | Claude 3.7 Sonnet | 58.4 | Feb 2025 | Jun 2025 | Official leaderboard | |
| 5 | Claude 3.5 Sonnet | 56.5 | Jun 2024 | Jun 2025 | Official leaderboard | |
| 6 | | 54.5 | Jan 2025 | Jun 2025 | Official leaderboard | |
| 7 | | 54.1 | Mar 2025 | Jun 2025 | Official leaderboard | |
| 8 | | 50.5 | Mar 2025 | Jun 2025 | Official leaderboard | |
| 9 | | 50.2 | Apr 2025 | Jun 2025 | Official leaderboard | |
| 10 | GPT-4o mini | 50.0 | Jul 2024 | Jun 2025 | Official leaderboard | |
| 11 | | 48.0 | Apr 2025 | Jun 2025 | Official leaderboard | |
| 12 | | 47.5 | Apr 2025 | Jun 2025 | Official leaderboard | |
| 13 | | 46.8 | Mar 2025 | Jun 2025 | Official leaderboard | |
| 14 | | 44.9 | Jan 2025 | Jun 2025 | Official leaderboard | |
| 15 | Claude 3 Haiku | 44.8 | Mar 2024 | Jun 2025 | Official leaderboard | |
| 16 | | 44.5 | Apr 2025 | Jun 2025 | Official leaderboard | |
| 17 | | 41.2 | Apr 2025 | Jun 2025 | Official leaderboard | |
| 18 | | 40.5 | Mar 2025 | Jun 2025 | Official leaderboard | |
| 19 | | 38.0 | Jan 2025 | Jun 2025 | Official leaderboard | |
| 20 | | 37.8 | Aug 2024 | Jun 2025 | Official leaderboard | |
| 21 | | 36.3 | Apr 2025 | Jun 2025 | Official leaderboard | |
| 22 | llava-v1.6-mistral-7b | 35.4 | Jan 2024 | Jun 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 Identification weighted score, Management weighted score and Identification accuracy. GPT-4.1 leads every one of them. Arrows show how far a model moved from its overall rank among the models we track.
Identification weighted score
22 models we track have a score here.
- 1 GPT-4.1 54.4
- 2 GPT-4o Mar 2025 ↑1 47.0
- 3 GPT-4.1 mini ↓1 46.0
- 4 Claude 3.7 Sonnet 44.6
- 5 Claude 3.5 Sonnet 42.2
- 6
Qwen2.5-VL-72B 40.5
- 7
Qwen2.5-VL-32B 37.0
- 8
InternVL3-78B ↑1 33.6
- 9 GPT-4o mini ↑1 33.1
- 10
Gemma 3 27B ↓2 31.1
Management weighted score
22 models we track have a score here.
- 1 GPT-4.1 81.6
- 2 GPT-4.1 mini 75.3
- 3 Claude 3.7 Sonnet ↑1 72.2
- 4
Qwen2.5-VL-32B ↑3 71.2
- 5 Claude 3.5 Sonnet 70.8
- 6 GPT-4o Mar 2025 ↓3 70.6
- 7
Gemma 3 27B ↑1 69.9
- 8
Qwen2.5-VL-72B ↓2 68.5
- 9 GPT-4o mini ↑1 66.9
- 10
InternVL3-78B ↓1 66.8
Identification accuracy
22 models we track have a score here.
- 1 GPT-4.1 43.9
- 2 GPT-4o Mar 2025 ↑1 39.3
- 3 GPT-4.1 mini ↓1 34.6
- 4 Claude 3.7 Sonnet 33.9
- 5 Claude 3.5 Sonnet 32.0
- 6
Qwen2.5-VL-72B 29.8
- 7
Qwen2.5-VL-32B 25.1
- 8 GPT-4o mini ↑2 22.4
- 9
InternVL3-78B 22.4
- 10
Qwen2.5-VL-7B ↑4 22.1
Identification reasoning
22 models we track have a score here.
- 1 GPT-4.1 3.01
- 2 GPT-4.1 mini 2.75
- 3 Claude 3.7 Sonnet ↑1 2.64
- 4 Claude 3.5 Sonnet ↑1 2.51
- 5 GPT-4o Mar 2025 ↓2 2.49
- 6
Qwen2.5-VL-72B 2.47
- 7
Qwen2.5-VL-32B 2.43
- 8
InternVL3-78B ↑1 2.24
- 9
Gemma 3 27B ↓1 2.22
- 10 GPT-4o mini 2.18
Management accuracy
22 models we track have a score here.
- 1 GPT-4.1 3.24
- 2 GPT-4.1 mini 2.94
- 3
Qwen2.5-VL-32B ↑4 2.87
- 4 Claude 3.7 Sonnet 2.82
- 5
Gemma 3 27B ↑3 2.77
- 6 GPT-4o Mar 2025 ↓3 2.77
- 7 Claude 3.5 Sonnet ↓2 2.75
- 8
Qwen2.5-VL-72B ↓2 2.72
- 9 GPT-4o mini ↑1 2.65
- 10
Gemma 3 12B ↑3 2.63
Management relevance
22 models we track have a score here.
- 1 GPT-4.1 3.60
- 2 GPT-4.1 mini 3.37
- 3 Claude 3.7 Sonnet ↑1 3.23
- 4
Qwen2.5-VL-32B ↑3 3.19
- 5 Claude 3.5 Sonnet 3.17
- 6 GPT-4o Mar 2025 ↓3 3.16
- 7
Gemma 3 27B ↑1 3.14
- 8
Qwen2.5-VL-72B ↓2 3.09
- 9 GPT-4o mini ↑1 3.02
- 10
Gemma 3 12B ↑3 3.00
Management completeness
22 models we track have a score here.
- 1 GPT-4.1 3.22
- 2
Qwen2.5-VL-32B ↑5 2.88
- 3
Gemma 3 27B ↑5 2.87
- 4 GPT-4.1 mini ↓2 2.83
- 5
Gemma 3 12B ↑8 2.74
- 6 Claude 3.7 Sonnet ↓2 2.69
- 7 Claude 3.5 Sonnet ↓2 2.57
- 8
Qwen2.5-VL-72B ↓2 2.56
- 9 GPT-4o Mar 2025 ↓6 2.43
- 10
Gemma 3 4B ↑8 2.32
Management parsimony
22 models we track have a score here.
- 1 GPT-4.1 3.01
- 2 GPT-4o Mar 2025 ↑1 3.00
- 3 GPT-4.1 mini ↓1 2.98
- 4 Claude 3.5 Sonnet ↑1 2.92
- 5 Claude 3.7 Sonnet ↓1 2.88
- 6
InternVL3-78B ↑3 2.87
- 7 Claude 3 Haiku ↑8 2.84
- 8 GPT-4o mini ↑2 2.78
- 9
InternVL3-38B ↑2 2.75
- 10
InternVL3-14B ↑6 2.66
Number one over time
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: Official MIRAGE-MMST 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 MIRAGE-MMST 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 Vision
- MMMU In use 11.5K college-level multimodal questions across 30 subjects. Multiple-choice and open accuracy.
Qwen 3.6 Plus 86%
- MMMU-Pro In use Harder MMMU: college-level multimodal questions with more options and vision-only inputs to remove text shortcuts. GPT-5.6 Sol 84.6%
- CharXiv In use Chart understanding on real arXiv figures. Reasoning questions require synthesizing information from complex charts. Claude Fable 5 93.5%
- MathVista In use Math reasoning in visual contexts (charts, geometry, puzzles) from 31 datasets. Answer accuracy.
Seed2.1 Pro 90.7%
- MATH-Vision (MathVision) In use 3,040 competition math problems with visual context. Answer accuracy.
Qwen 3.8 Max 97.7%
- MMStar In use Models answer 1,500 human-curated image questions that cannot be solved without the image, across 6 core capabilities. Scored by accuracy.
Qwen 3.8 Max 85.9%
Data audit
8 of 9 checks pass, 1 warning. Look first at the Newest score check. The newest came 19 months ago, in April 2025, for GPT-4.1. 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 (22 entries).
- Pass Models scored 22 models we track.
- Warning Newest score The newest came 19 months ago, in April 2025, for GPT-4.1.
- Note Frontier models None of the top 10 has a score. The strongest tested is Claude 3.7 Sonnet (#119).
- Note Headroom Scored as a plain number, with no fixed ceiling.
Leaderboard import 3 pass
- Pass Importer We read the board on mirage-benchmark.github.io: 22 rows for models we track 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.