EarthVerse
Agents investigate documented natural hazards using event data from different sources. The evaluator checks individual claims and the calculations that support them across 405 tasks.
- Made by
- Zhiqing Cui and colleagues
- Released
- August 2026
- Kind
- Lab and data work
- Level
- Expert
- Scored
- Checked automatically
- Answers
- Agent task
- Measures
- Capability
- Subject
- Earth and environmental sciences
- Language
- English
- In and out
- Text, Maps and geodata, Time series → Actions, Text
- Interaction
- Agentic
- Environment
- Code sandbox
- Items from
- Real work and use
- Refresh
- One fixed set
- Run by
- Anyone
- In views
- Agents
- Strongest model tested
- Claude Fable 5 #12 on the capability index
- Last model added
- July 2026
- Scores refreshed
- November 20th, 2026
Claude Fable 5 leads with 85% among models released by July 2026 Claude Fable 5 leads with 85% among models released by July 2026 GLM-5.2 leads with 79.2% in EarthVerse
Best score per entry from the benchmark's own leaderboard, read on November 20th, 2026. Mean core averages answer correctness and the rubric score. Where a model ran on its own with several settings, we keep its best. Every run is in the next chart. Agents other companies built around a model get a pale shade of that model's lab color. Gray marks blends and entries not in our model list, such as fine-tunes and older models. None of these count toward leaders. Agents built around a model, such as GeoMMAgent + GPT-5.5, are their own entries. We don't count them as the model's score. This view shows only models we track, each at its best. This view keeps only runs in EarthVerse, a harness no model’s maker built. A maker’s own agent, such as Claude Code with a Claude model, is left out.
The newest model with a score here came out in July 2026, 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 13 more
- 85.0 Claude Fable 5
- 81.7 GPT-5.6 Sol
- 80.1 GPT-5.6 Terra
- 79.2 GLM-5.2
- 78.1 GPT-5.5
- 76.2 Claude Sonnet 4.6
- 75.3 Kimi K2.5
- 73.5
DeepSeek-V4-Flash
- 72.8
Tencent Hy3
- 72.2 Claude Haiku 4.5
- 70.3
Qwen3.7-Plus
- 68.3 GeoMMAgent + GPT-5.5
- 63.0
Seed2.0 Mini
- 59.7
Gemini 3.1 Flash-Lite
- 44.7
MiniMax-M2.7
- 39.2 GPT-4o (snapshot unspecified)
- 35.7 GPT-5.6 Luna
- 33.3 Qwen3-235B
- 31.1 GeohazardGPT 8B
- 30.8 OpenResearcher
- 25.0
Qwen3-4B
- 24.7 GeoGPT-SFT 4B
- 13.2 MiroThinker 8B
- 13.2 nvidia-nemotron-nano-9b-v2
- 11.7 Intern-S1-mini
- 1 Claude Fable 5 85%
- 2 GPT-5.6 Sol 81.7%
- 3 GPT-5.6 Terra 80.1%
- 4 GLM-5.2 79.2%
- 5 GPT-5.5 78.1%
- 6 Claude Sonnet 4.6 76.2%
- 7 Kimi K2.5 75.3%
- 8
DeepSeek-V4-Flash 73.5%
- 9
Tencent Hy3 72.8%
- 10 Claude Haiku 4.5 72.2%
- 11
Qwen3.7-Plus 70.3%
- 12 GeoMMAgent + GPT-5.5 68.3%
- 13
Seed2.0 Mini 63%
- 14
Gemini 3.1 Flash-Lite 59.7%
- 15
MiniMax-M2.7 44.7%
- 16 GPT-4o (snapshot unspecified) 39.2%
- 17 GPT-5.6 Luna 35.7%
- 18 Qwen3-235B 33.3%
- 19 GeohazardGPT 8B 31.1%
- 20 OpenResearcher 30.8%
- 21
Qwen3-4B 25%
- 22 GeoGPT-SFT 4B 24.7%
- 23 MiroThinker 8B 13.2%
- 24 nvidia-nemotron-nano-9b-v2 13.2%
- 25 Intern-S1-mini 11.7%
Source: Official EarthVerse leaderboard
- 85.0 Claude Fable 5
- 81.7 GPT-5.6 Sol
- 80.1 GPT-5.6 Terra
- 79.2 GLM-5.2
- 78.1 GPT-5.5
- 76.2 Claude Sonnet 4.6
- 75.3 Kimi K2.5
- 73.5
DeepSeek-V4-Flash
- 72.8
Tencent Hy3
- 72.2 Claude Haiku 4.5
- 70.3
Qwen3.7-Plus
- 63.0
Seed2.0 Mini
- 59.7
Gemini 3.1 Flash-Lite
- 44.7
MiniMax-M2.7
- 35.7 GPT-5.6 Luna
- 25.0
Qwen3-4B
- 13.2 nvidia-nemotron-nano-9b-v2
- 1 Claude Fable 5 85%
- 2 GPT-5.6 Sol 81.7%
- 3 GPT-5.6 Terra 80.1%
- 4 GLM-5.2 79.2%
- 5 GPT-5.5 78.1%
- 6 Claude Sonnet 4.6 76.2%
- 7 Kimi K2.5 75.3%
- 8
DeepSeek-V4-Flash 73.5%
- 9
Tencent Hy3 72.8%
- 10 Claude Haiku 4.5 72.2%
- 11
Qwen3.7-Plus 70.3%
- 12
Seed2.0 Mini 63%
- 13
Gemini 3.1 Flash-Lite 59.7%
- 14
MiniMax-M2.7 44.7%
- 15 GPT-5.6 Luna 35.7%
- 16
Qwen3-4B 25%
- 17 nvidia-nemotron-nano-9b-v2 13.2%
Source: Official EarthVerse leaderboard
- 1 GLM-5.2 79.2%
- 2 GPT-5.5 78.1%
- 3 Claude Sonnet 4.6 76.2%
- 4 Kimi K2.5 75.3%
- 5
DeepSeek-V4-Flash 73.5%
- 6
Tencent Hy3 72.8%
- 7 Claude Haiku 4.5 72.2%
- 8
Qwen3.7-Plus 70.3%
- 9
Seed2.0 Mini 63%
- 10
Gemini 3.1 Flash-Lite 59.7%
- 11
MiniMax-M2.7 44.7%
- 12
Qwen3-4B 25%
- 13 nvidia-nemotron-nano-9b-v2 13.2%
Source: Official EarthVerse leaderboard
| # | Entry | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | Claude Fable 5 | 85% | Claude Code | Jun 2026 | Official leaderboard |
| 2 | GPT-5.6 Sol | 81.7% | Codex | Jul 2026 | Official leaderboard |
| 3 | GPT-5.6 Terra | 80.1% | Codex | Jul 2026 | Official leaderboard |
| 4 | GLM-5.2 | 79.2% | EarthVerse | Jun 2026 | Official leaderboard |
| 5 | GPT-5.5 | 78.1% | EarthVerse | Apr 2026 | Official leaderboard |
| 6 | Claude Sonnet 4.6 | 76.2% | EarthVerse | Feb 2026 | Official leaderboard |
| 7 | Kimi K2.5 | 75.3% | EarthVerse | Jan 2026 | Official leaderboard |
| 8 | | 73.5% | EarthVerse | Apr 2026 | Official leaderboard |
| 9 | | 72.8% | EarthVerse | Jul 2026 | Official leaderboard |
| 10 | Claude Haiku 4.5 | 72.2% | EarthVerse | Oct 2025 | Official leaderboard |
| 11 | | 70.3% | EarthVerse | Jun 2026 | Official leaderboard |
| 12 | GeoMMAgent + GPT-5.5 | 68.3% | Apr 2026 | Official leaderboard | |
| 13 | | 63% | EarthVerse | Feb 2026 | Official leaderboard |
| 14 | | 59.7% | EarthVerse | Mar 2026 | Official leaderboard |
| 15 | | 44.7% | EarthVerse | Mar 2026 | Official leaderboard |
| 16 | GPT-4o (snapshot unspecified) | 39.2% | EarthVerse | Official leaderboard | |
| 17 | GPT-5.6 Luna | 35.7% | Codex | Jul 2026 | Official leaderboard |
| 18 | Qwen3-235B | 33.3% | EarthVerse | Official leaderboard | |
| 19 | GeohazardGPT 8B | 31.1% | EarthVerse | Official leaderboard | |
| 20 | OpenResearcher | 30.8% | OpenResearcher | Official leaderboard | |
| 21 | | 25% | EarthVerse | Apr 2025 | Official leaderboard |
| 22 | GeoGPT-SFT 4B | 24.7% | EarthVerse | Official leaderboard | |
| 23 | MiroThinker 8B | 13.2% | EarthVerse | Official leaderboard | |
| 24 | nvidia-nemotron-nano-9b-v2 | 13.2% | EarthVerse | Aug 2025 | Official leaderboard |
| 25 | Intern-S1-mini | 11.7% | EarthVerse | Official leaderboard |
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | Claude Fable 5 | 85% | Claude Code | Jun 2026 | Official leaderboard |
| 2 | GPT-5.6 Sol | 81.7% | Codex | Jul 2026 | Official leaderboard |
| 3 | GPT-5.6 Terra | 80.1% | Codex | Jul 2026 | Official leaderboard |
| 4 | GLM-5.2 | 79.2% | EarthVerse | Jun 2026 | Official leaderboard |
| 5 | GPT-5.5 | 78.1% | EarthVerse | Apr 2026 | Official leaderboard |
| 6 | Claude Sonnet 4.6 | 76.2% | EarthVerse | Feb 2026 | Official leaderboard |
| 7 | Kimi K2.5 | 75.3% | EarthVerse | Jan 2026 | Official leaderboard |
| 8 | | 73.5% | EarthVerse | Apr 2026 | Official leaderboard |
| 9 | | 72.8% | EarthVerse | Jul 2026 | Official leaderboard |
| 10 | Claude Haiku 4.5 | 72.2% | EarthVerse | Oct 2025 | Official leaderboard |
| 11 | | 70.3% | EarthVerse | Jun 2026 | Official leaderboard |
| 12 | | 63% | EarthVerse | Feb 2026 | Official leaderboard |
| 13 | | 59.7% | EarthVerse | Mar 2026 | Official leaderboard |
| 14 | | 44.7% | EarthVerse | Mar 2026 | Official leaderboard |
| 15 | GPT-5.6 Luna | 35.7% | Codex | Jul 2026 | Official leaderboard |
| 16 | | 25% | EarthVerse | Apr 2025 | Official leaderboard |
| 17 | nvidia-nemotron-nano-9b-v2 | 13.2% | EarthVerse | Aug 2025 | Official leaderboard |
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | GLM-5.2 | 79.2% | EarthVerse | Jun 2026 | Official leaderboard |
| 2 | GPT-5.5 | 78.1% | EarthVerse | Apr 2026 | Official leaderboard |
| 3 | Claude Sonnet 4.6 | 76.2% | EarthVerse | Feb 2026 | Official leaderboard |
| 4 | Kimi K2.5 | 75.3% | EarthVerse | Jan 2026 | Official leaderboard |
| 5 | | 73.5% | EarthVerse | Apr 2026 | Official leaderboard |
| 6 | | 72.8% | EarthVerse | Jul 2026 | Official leaderboard |
| 7 | Claude Haiku 4.5 | 72.2% | EarthVerse | Oct 2025 | Official leaderboard |
| 8 | | 70.3% | EarthVerse | Jun 2026 | Official leaderboard |
| 9 | | 63% | EarthVerse | Feb 2026 | Official leaderboard |
| 10 | | 59.7% | EarthVerse | Mar 2026 | Official leaderboard |
| 11 | | 44.7% | EarthVerse | Mar 2026 | Official leaderboard |
| 12 | | 25% | EarthVerse | Apr 2025 | Official leaderboard |
| 13 | nvidia-nemotron-nano-9b-v2 | 13.2% | EarthVerse | Aug 2025 | Official leaderboard |
Rankings by part
We read 10 parts from the benchmark’s own leaderboard, each ranking models on one part of the test, such as L2, L3 and L4. Claude Fable 5 leads 9 of them, and GPT-5.6 Sol leads the other one. Arrows show how far a model moved from its overall rank among the models we track.
L2
17 models we track have a score here.
- 1 GPT-5.6 Sol ↑1 88%
- 2 Claude Fable 5 ↓1 86.6%
- 3 GPT-5.6 Terra 83.2%
- 4 GPT-5.5 ↑1 81.6%
- 5 GLM-5.2 ↓1 80.5%
- 6 Kimi K2.5 ↑1 79%
- 7
DeepSeek-V4-Flash ↑1 75.9%
- 8 Claude Sonnet 4.6 ↓2 75.8%
- 9 Claude Haiku 4.5 ↑1 73.7%
- 10
Qwen3.7-Plus ↑1 73.1%
L3
17 models we track have a score here.
- 1 Claude Fable 5 86.5%
- 2 GLM-5.2 ↑2 83.3%
- 3 GPT-5.5 ↑2 81.4%
- 4 Claude Sonnet 4.6 ↑2 80.2%
- 5 GPT-5.6 Sol ↓3 80%
- 6 GPT-5.6 Terra ↓3 79.2%
- 7 Claude Haiku 4.5 ↑3 78%
- 8
Tencent Hy3 ↑1 76%
- 9
DeepSeek-V4-Flash ↓1 74.8%
- 10 Kimi K2.5 ↓3 74.2%
L4
17 models we track have a score here.
- 1 Claude Fable 5 80.4%
- 2 Claude Sonnet 4.6 ↑4 76.4%
- 3 GPT-5.5 ↑2 75.9%
- 4 GPT-5.6 Sol ↓2 75.8%
- 5 GPT-5.6 Terra ↓2 75.6%
- 6 Kimi K2.5 ↑1 74%
- 7 GLM-5.2 ↓3 72.2%
- 8
Tencent Hy3 ↑1 70.5%
- 9
Qwen3.7-Plus ↑2 69.9%
- 10
DeepSeek-V4-Flash ↓2 69.2%
Physical mechanism
17 models we track have a score here.
- 1 Claude Fable 5 85.2%
- 2 GPT-5.6 Sol 81.8%
- 3 GPT-5.6 Terra 80.8%
- 4 GLM-5.2 80.3%
- 5 GPT-5.5 77.2%
- 6 Claude Sonnet 4.6 75.5%
- 7 Kimi K2.5 75.1%
- 8
DeepSeek-V4-Flash 73.6%
- 9
Tencent Hy3 72.7%
- 10
Qwen3.7-Plus ↑1 70.4%
Spatiotemporal process
17 models we track have a score here.
- 1 Claude Fable 5 85.5%
- 2 GPT-5.6 Sol 83.3%
- 3 GPT-5.6 Terra 79.8%
- 4 Kimi K2.5 ↑3 79.1%
- 5 GPT-5.5 78.4%
- 6 GLM-5.2 ↓2 77.3%
- 7
DeepSeek-V4-Flash ↑1 76.3%
- 8
Tencent Hy3 ↑1 75.9%
- 9 Claude Sonnet 4.6 ↓3 73.6%
- 10 Claude Haiku 4.5 72.4%
Quantitative calculation
17 models we track have a score here.
- 1 Claude Fable 5 87.1%
- 2 GPT-5.6 Sol 85.4%
- 3 GPT-5.6 Terra 83%
- 4 GPT-5.5 ↑1 81.9%
- 5 GLM-5.2 ↓1 80.2%
- 6 Claude Sonnet 4.6 78.4%
- 7 Kimi K2.5 77.8%
- 8 Claude Haiku 4.5 ↑2 76.1%
- 9
DeepSeek-V4-Flash ↓1 75.2%
- 10
Tencent Hy3 ↓1 73.7%
Multi-source evidence
17 models we track have a score here.
- 1 Claude Fable 5 83.5%
- 2 GPT-5.6 Sol 82.1%
- 3 GPT-5.6 Terra 79.1%
- 4 Claude Sonnet 4.6 ↑2 77.9%
- 5 GLM-5.2 ↓1 76.8%
- 6 GPT-5.5 ↓1 76.8%
- 7 Kimi K2.5 75.3%
- 8 Claude Haiku 4.5 ↑2 73.5%
- 9
DeepSeek-V4-Flash ↓1 72.6%
- 10
Tencent Hy3 ↓1 70.3%
Causal chain
17 models we track have a score here.
- 1 Claude Fable 5 82.2%
- 2 GLM-5.2 ↑2 80.1%
- 3 GPT-5.6 Sol ↓1 77.5%
- 4 GPT-5.6 Terra ↓1 76.3%
- 5 GPT-5.5 74.3%
- 6 Claude Sonnet 4.6 73.4%
- 7 Kimi K2.5 70.5%
- 8
Tencent Hy3 ↑1 69.7%
- 9
DeepSeek-V4-Flash ↓1 68.9%
- 10 Claude Haiku 4.5 67.7%
Ranking decision
17 models we track have a score here.
- 1 Claude Fable 5 81%
- 2 Claude Sonnet 4.6 ↑4 79.7%
- 3 GLM-5.2 ↑1 78.3%
- 4 GPT-5.6 Terra ↓1 76.1%
- 5 GPT-5.5 75.8%
- 6
DeepSeek-V4-Flash ↑2 72.3%
- 7 GPT-5.6 Sol ↓5 71.6%
- 8 Kimi K2.5 ↓1 70.8%
- 9
Tencent Hy3 70.2%
- 10
Qwen3.7-Plus ↑1 68.1%
Remote sensing and geospatial
17 models we track have a score here.
- 1 Claude Fable 5 90%
- 2 GPT-5.6 Sol 88.5%
- 3 GPT-5.6 Terra 86.3%
- 4 GPT-5.5 ↑1 83.8%
- 5 Claude Haiku 4.5 ↑5 79.7%
- 6
Tencent Hy3 ↑3 79.4%
- 7 GLM-5.2 ↓3 79.3%
- 8
Qwen3.7-Plus ↑3 78.3%
- 9 Kimi K2.5 ↓2 77.9%
- 10
DeepSeek-V4-Flash ↓2 77.8%
The best score, 85%, costs $3,272 for the full run
Runs on the cost frontier score higher than any run that's cheaper. GPT-5.6 Terra comes within 4.9 points of the best for $1,109 for the full run, 3 times cheaper. The costs are the leaderboard's own estimates, so compare them only within this benchmark.
Source: Official EarthVerse leaderboard
| Model | Effort | Harness | Score | Cost of the full run | Output tokens |
|---|---|---|---|---|---|
| Claude Fable 5 | Claude Code | 85% | $3,272 | ||
| GPT-5.6 Sol | Codex | 81.7% | $2,502 | ||
| GPT-5.6 Terra | Codex | 80.1% | $1,109 | ||
| GLM-5.2 | EarthVerse | 79.2% | |||
| GPT-5.5 | EarthVerse | 78.1% | $2,732 | ||
| Claude Sonnet 4.6 | EarthVerse | 76.2% | $2,011 | ||
| Kimi K2.5 | EarthVerse | 75.3% | |||
| EarthVerse | 73.5% | $44 | |||
| EarthVerse | 72.8% | ||||
| Claude Haiku 4.5 | EarthVerse | 72.2% | $966 | ||
| EarthVerse | 70.3% | ||||
| GeoMMAgent + GPT-5.5 | GeoMMAgent | 68.3% | |||
| EarthVerse | 63% | ||||
| EarthVerse | 59.7% | $43 | |||
| EarthVerse | 44.7% | ||||
| GPT-5.6 Luna | Codex | 35.7% | $381 | ||
| EarthVerse | 25% | ||||
| nvidia-nemotron-nano-9b-v2 | EarthVerse | 13.2% |
The best score, 85%, costs $3,272 for the full run
Runs on the cost frontier score higher than any run that's cheaper. GPT-5.6 Terra comes within 4.9 points of the best for $1,109 for the full run, 3 times cheaper. The costs are the leaderboard's own estimates, so compare them only within this benchmark.
Source: Official EarthVerse leaderboard
| Model | Effort | Harness | Score | Cost of the full run | Output tokens |
|---|---|---|---|---|---|
| Claude Fable 5 | Claude Code | 85% | $3,272 | ||
| GPT-5.6 Sol | Codex | 81.7% | $2,502 | ||
| GPT-5.6 Terra | Codex | 80.1% | $1,109 | ||
| GLM-5.2 | EarthVerse | 79.2% | |||
| GPT-5.5 | EarthVerse | 78.1% | $2,732 | ||
| Claude Sonnet 4.6 | EarthVerse | 76.2% | $2,011 | ||
| Kimi K2.5 | EarthVerse | 75.3% | |||
| EarthVerse | 73.5% | $44 | |||
| EarthVerse | 72.8% | ||||
| Claude Haiku 4.5 | EarthVerse | 72.2% | $966 | ||
| EarthVerse | 70.3% | ||||
| EarthVerse | 63% | ||||
| EarthVerse | 59.7% | $43 | |||
| EarthVerse | 44.7% | ||||
| GPT-5.6 Luna | Codex | 35.7% | $381 | ||
| EarthVerse | 25% | ||||
| nvidia-nemotron-nano-9b-v2 | EarthVerse | 13.2% |
Number one over time
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: Official EarthVerse 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 EarthVerse leaderboard
Source: Official EarthVerse 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 Task access check. We don’t know whether the tasks are public. Every benchmark
Catalog record 1 warning 1 pass
- Warning Task access We don’t know whether the tasks are public.
- Pass Confidence Checked against a primary source.
Scores 4 pass 1 note
- Pass Independent runs Scores from the official leaderboard (25 entries).
- Pass Models scored 17 models we track, plus 8 other entries on its leaderboard.
- Pass Newest score The newest is from July 2026, for GPT-5.6 Luna.
- Note Frontier models None of the top 10 has a score. The strongest tested is Claude Fable 5 (#12).
- Pass Headroom The best is 85%, under 85%, so it still separates models.
Leaderboard import 3 pass
- Pass Importer We read the board on cuizhiq.github.io: 18 rows for models we track and 7 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.