OmniEarth-Bench
Models interpret satellite images and observational data across Earth systems, answering questions about environmental processes and their interactions. Multiple-choice accuracy and model judgments score answers. Separate measures check localization, captions, and annotated reasoning steps.
The initial release has 29,779 annotations. The site and revised paper give different task counts, so this record does not equate their dimension totals. Scores for answer selection, localization, captions, and reasoning are not interchangeable.
- Released
- May 2025
- Kind
- Research science
- Level
- Expert
- Scored
- Graded by a model
- Answers
- Open-ended
- Tasks
- All public
- Measures
- Capability
- Subject
- Earth and environmental sciences
- Language
- English
- In and out
- Text, Images, Maps and geodata → Text
- 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
- March 2025
- Scores refreshed
- November 18th, 2026
In the authors’ words
Existing benchmarks for multimodal learning in Earth science offer limited, siloed coverage of Earth's spheres and their cross-sphere interactions, typically restricting evaluation to the human-activity sphere of atmosphere and to at most 16 tasks. These limitations: narrow-source heterogeneity (single/few data sources), constrained scientific granularity, and limited-sphere extensibility. Therefore, we introduce OmniEarth-Bench, the first multimodal benchmark that systematically spans all six spheres: atmosphere, lithosphere, oceanosphere, cryosphere, biosphere, and human-activity sphere, and cross-spheres. Built with a scalable, modular-topology data inference framework and native multi-observation sources and expert-in-the-loop curation, OmniEarth-Bench produces 29,855 standardized, expert-curated annotations. All annotations are organized into a four-level hierarchy (Sphere, Scenario, Ability, Task), encompassing 109 expert-curated evaluation tasks. Experiments on 9 state-of-the-art MLLMs reveal that even the most advanced models struggle with our benchmarks, where none of them reach 35% accuracy, revealing systematic gaps in Earth-system cognitive ability. The dataset and evaluation code were released at OmniEarth-Bench (https://anonymous.4open.science/r/OmniEarth-Bench-B1BD).
InternVL3-72B tops the leaderboard with 33.3% LLaVA-OneVision 7B leads with 31.5% among models released by March 2025
Best score per entry from the benchmark's own leaderboard, read on November 18th, 2026. 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. The best model we track is LLaVA-OneVision 7B, at 31.5%. This view shows only models we track, each at its best.
The newest model with a score here came out in March 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
- 1 InternVL3-72B 33.3%
- 2 InternVL3-7B 33.1%
- 3
LLaVA-OneVision 7B 31.5%
- 4 Claude 3.7 Sonnet 29.1%
- 5 Gemini-2.0 28.1%
- 6
InternLM-XComposer2.5-7B 26.1%
- 7
Qwen2.5-VL-7B 12.4%
- 8 GPT-4o Mar 2025 11.2%
- 9
Qwen2.5-VL-72B 11%
Source: Official OmniEarth-Bench leaderboard
- 1
LLaVA-OneVision 7B 31.5%
- 2 Claude 3.7 Sonnet 29.1%
- 3
InternLM-XComposer2.5-7B 26.1%
- 4
Qwen2.5-VL-7B 12.4%
- 5 GPT-4o Mar 2025 11.2%
- 6
Qwen2.5-VL-72B 11%
Source: Official OmniEarth-Bench leaderboard
| # | Entry | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | InternVL3-72B | 33.3% | Official leaderboard | ||
| 2 | InternVL3-7B | 33.1% | Official leaderboard | ||
| 3 | | 31.5% | Aug 2024 | Official leaderboard | |
| 4 | Claude 3.7 Sonnet | 29.1% | Feb 2025 | Official leaderboard | |
| 5 | Gemini-2.0 | 28.1% | Official leaderboard | ||
| 6 | | 26.1% | Jul 2024 | Official leaderboard | |
| 7 | | 12.4% | Jan 2025 | Official leaderboard | |
| 8 | GPT-4o Mar 2025 | 11.2% | Mar 2025 | Official leaderboard | |
| 9 | | 11% | Jan 2025 | Official leaderboard |
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | | 31.5% | Aug 2024 | Official leaderboard | |
| 2 | Claude 3.7 Sonnet | 29.1% | Feb 2025 | Official leaderboard | |
| 3 | | 26.1% | Jul 2024 | Official leaderboard | |
| 4 | | 12.4% | Jan 2025 | Official leaderboard | |
| 5 | GPT-4o Mar 2025 | 11.2% | Mar 2025 | Official leaderboard | |
| 6 | | 11% | Jan 2025 | Official leaderboard |
Rankings by part
We read 7 parts from the benchmark’s own leaderboard, each ranking models on one part of the test, such as Cross sphere, Atmosphere and Lithosphere. LLaVA-OneVision 7B leads 4 of them, and 2 other models lead the rest. Arrows show how far a model moved from its overall rank among the models we track.
Cross sphere
6 models we track have a score here.
- 1 Claude 3.7 Sonnet ↑1 30.7%
- 2
InternLM-XComposer2.5-7B ↑1 19.8%
- 3
LLaVA-OneVision 7B ↓2 19.3%
- 4
Qwen2.5-VL-7B 9.85%
- 5
Qwen2.5-VL-72B ↑1 3.92%
- 6 GPT-4o Mar 2025 ↓1 0.04%
Atmosphere
6 models we track have a score here.
- 1
LLaVA-OneVision 7B 33.7%
- 2 Claude 3.7 Sonnet 24.7%
- 3
InternLM-XComposer2.5-7B 17.4%
- 4 GPT-4o Mar 2025 ↑1 9.64%
- 5
Qwen2.5-VL-7B ↓1 9.25%
- 6
Qwen2.5-VL-72B 4.82%
Lithosphere
6 models we track have a score here.
- 1
InternLM-XComposer2.5-7B ↑2 28.9%
- 2
LLaVA-OneVision 7B ↓1 28.7%
- 3 Claude 3.7 Sonnet ↓1 28.1%
- 4
Qwen2.5-VL-72B ↑2 22.4%
- 5
Qwen2.5-VL-7B ↓1 18.6%
- 6 GPT-4o Mar 2025 ↓1 12.8%
Oceansphere
6 models we track have a score here.
- 1
LLaVA-OneVision 7B 24.5%
- 2 Claude 3.7 Sonnet 23.1%
- 3
InternLM-XComposer2.5-7B 21.1%
- 4
Qwen2.5-VL-72B ↑2 16.3%
- 5
Qwen2.5-VL-7B ↓1 13.9%
- 6 GPT-4o Mar 2025 ↓1 13.3%
Cryosphere
6 models we track have a score here.
- 1 Claude 3.7 Sonnet ↑1 54.5%
- 2
LLaVA-OneVision 7B ↓1 46.4%
- 3
InternLM-XComposer2.5-7B 40%
- 4 GPT-4o Mar 2025 ↑1 37.5%
- 5
Qwen2.5-VL-7B ↓1 17.9%
- 6
Qwen2.5-VL-72B 5.88%
Biosphere
6 models we track have a score here.
- 1
LLaVA-OneVision 7B 37.3%
- 2 Claude 3.7 Sonnet 31.2%
- 3
InternLM-XComposer2.5-7B 30.7%
- 4
Qwen2.5-VL-72B ↑2 14.9%
- 5
Qwen2.5-VL-7B ↓1 10.9%
- 6 GPT-4o Mar 2025 ↓1 1.97%
Human activities
6 models we track have a score here.
- 1
LLaVA-OneVision 7B 30.6%
- 2
InternLM-XComposer2.5-7B ↑1 24.8%
- 3 Claude 3.7 Sonnet ↓1 11.2%
- 4
Qwen2.5-VL-72B ↑2 8.63%
- 5
Qwen2.5-VL-7B ↓1 6.23%
- 6 GPT-4o Mar 2025 ↓1 2.76%
Number one over time
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: Official OmniEarth-Bench 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 OmniEarth-Bench 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 Science and reasoning
- GPQA Diamond In use 198 graduate-level 'Google-proof' multiple-choice questions in biology, chemistry and physics.
Gemini 4 Carbon 97%
- Terminal-Bench-Science In use 70 scientific research-workflow tasks across five science areas, run in Harbor containers. Hidden tests give the resolution rate. GPT-6 Astra 68.1%
- BioMysteryBench In use Bioinformatics problems on anonymized real datasets that require real analysis with tools. Rubric-graded, with human-solvable and human-difficult splits. Claude Fable 5.2 83.7%
- FrontierScience In use Models solve physics, chemistry, and biology problems in an Olympiad track graded by short answer and a Research track graded by 10-point rubrics.
Intern-S2-397B 87%
- HiPhO In use Models solve theory problems from 2024 and 2025 physics olympiad exams, some with diagrams. Official marking schemes grade answers and steps, mapped to medal levels.
Gemini 3 Pro 78.3%
- LABBench2 In use 11 biology research sub-tasks (literature, figures, tables, patents, trials and more) with tool access. Macro-averaged accuracy. GPT-5.2 60.9%
Data audit
9 of 10 checks pass, 1 warning. Look first at the Newest score check. The newest came 20 months ago, in March 2025, for GPT-4o. Every benchmark
Catalog record 2 pass
- Pass Task access All tasks are public.
- Pass Confidence Checked against a primary source.
Scores 1 warning 3 pass 1 note
- Pass Independent runs Scores from the official leaderboard (9 entries).
- Pass Models scored 6 models we track, plus 3 other entries on its leaderboard.
- Warning Newest score The newest came 20 months ago, in March 2025, for GPT-4o.
- Note Frontier models None of the top 10 has a score. The strongest tested is Claude 3.7 Sonnet (#119).
- Pass Headroom The best is 31.5%, under 85%, so it still separates models.
Leaderboard import 3 pass
- Pass Importer We read the board on omniearth-bench.github.io: 6 rows for models we track and 3 other entries at the last read.
- Pass Last read November 18th, 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.