Benchmark Almanac

EarthVerse

Claude Fable 5 85% leads since August 2026 GLM-5.2 79.2% leads in EarthVerse
Not in model cards

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
Language
English
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
25
Entries on its own leaderboard, 17 models we track

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:

  1. 1 Claude Fable 5 85%
  2. 2 GPT-5.6 Sol 81.7%
  3. 3 GPT-5.6 Terra 80.1%
  4. 4 GLM-5.2 79.2%
  5. 5 GPT-5.5 78.1%
  6. 6 Claude Sonnet 4.6 76.2%
  7. 7 Kimi K2.5 75.3%
  8. 8 DeepSeek-V4-Flash 73.5%
  9. 9 Tencent Hy3 72.8%
  10. 10 Claude Haiku 4.5 72.2%
  11. 11 Qwen3.7-Plus 70.3%
  12. 12 GeoMMAgent + GPT-5.5 68.3%
  13. 13 Seed2.0 Mini 63%
  14. 14 Gemini 3.1 Flash-Lite 59.7%
  15. 15 MiniMax-M2.7 44.7%
  16. 16 GPT-4o (snapshot unspecified) 39.2%
  17. 17 GPT-5.6 Luna 35.7%
  18. 18 Qwen3-235B 33.3%
  19. 19 GeohazardGPT 8B 31.1%
  20. 20 OpenResearcher 30.8%
  21. 21 Qwen3-4B 25%
  22. 22 GeoGPT-SFT 4B 24.7%
  23. 23 MiroThinker 8B 13.2%
  24. 24 nvidia-nemotron-nano-9b-v2 13.2%
  25. 25 Intern-S1-mini 11.7%

Source: Official EarthVerse leaderboard

#EntryScoreSettingDateSource
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 DeepSeek-V4-Flash 73.5% EarthVerse Apr 2026 Official leaderboard
9 Tencent Hy3 72.8% EarthVerse Jul 2026 Official leaderboard
10 Claude Haiku 4.5 72.2% EarthVerse Oct 2025 Official leaderboard
11 Qwen3.7-Plus 70.3% EarthVerse Jun 2026 Official leaderboard
12 GeoMMAgent + GPT-5.5 68.3% Apr 2026 Official leaderboard
13 Seed2.0 Mini 63% EarthVerse Feb 2026 Official leaderboard
14 Gemini 3.1 Flash-Lite 59.7% EarthVerse Mar 2026 Official leaderboard
15 MiniMax-M2.7 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 Qwen3-4B 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
#ModelScoreSettingDateSource
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 DeepSeek-V4-Flash 73.5% EarthVerse Apr 2026 Official leaderboard
9 Tencent Hy3 72.8% EarthVerse Jul 2026 Official leaderboard
10 Claude Haiku 4.5 72.2% EarthVerse Oct 2025 Official leaderboard
11 Qwen3.7-Plus 70.3% EarthVerse Jun 2026 Official leaderboard
12 Seed2.0 Mini 63% EarthVerse Feb 2026 Official leaderboard
13 Gemini 3.1 Flash-Lite 59.7% EarthVerse Mar 2026 Official leaderboard
14 MiniMax-M2.7 44.7% EarthVerse Mar 2026 Official leaderboard
15 GPT-5.6 Luna 35.7% Codex Jul 2026 Official leaderboard
16 Qwen3-4B 25% EarthVerse Apr 2025 Official leaderboard
17 nvidia-nemotron-nano-9b-v2 13.2% EarthVerse Aug 2025 Official leaderboard
#ModelScoreSettingDateSource
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 DeepSeek-V4-Flash 73.5% EarthVerse Apr 2026 Official leaderboard
6 Tencent Hy3 72.8% EarthVerse Jul 2026 Official leaderboard
7 Claude Haiku 4.5 72.2% EarthVerse Oct 2025 Official leaderboard
8 Qwen3.7-Plus 70.3% EarthVerse Jun 2026 Official leaderboard
9 Seed2.0 Mini 63% EarthVerse Feb 2026 Official leaderboard
10 Gemini 3.1 Flash-Lite 59.7% EarthVerse Mar 2026 Official leaderboard
11 MiniMax-M2.7 44.7% EarthVerse Mar 2026 Official leaderboard
12 Qwen3-4B 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. 1 GPT-5.6 Sol ↑1 88%
  2. 2 Claude Fable 5 ↓1 86.6%
  3. 3 GPT-5.6 Terra 83.2%
  4. 4 GPT-5.5 ↑1 81.6%
  5. 5 GLM-5.2 ↓1 80.5%
  6. 6 Kimi K2.5 ↑1 79%
  7. 7 DeepSeek-V4-Flash ↑1 75.9%
  8. 8 Claude Sonnet 4.6 ↓2 75.8%
  9. 9 Claude Haiku 4.5 ↑1 73.7%
  10. 10 Qwen3.7-Plus ↑1 73.1%

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.

Google DeepMind DeepSeek Anthropic OpenAI Cost frontier One model at each effort level
0% 20% 40% 60% 80% 100% $10 $100 $1,000 $10,000 Cost of the full run, log scale Claude Fable 5 GPT-5.6 Terra DeepSeek-V4-Flash Gemini 3.1 Flash-Lite 0% 20% 40% 60% 80% 100% $10 $100 $1,000 $10,000 Cost of the full run, log scale Claude Fable 5 DeepSeek-V4-Flash

Source: Official EarthVerse leaderboard

ModelEffortHarnessScoreCost of the full runOutput 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%
DeepSeek-V4-Flash EarthVerse 73.5% $44
Tencent Hy3 EarthVerse 72.8%
Claude Haiku 4.5 EarthVerse 72.2% $966
Qwen3.7-Plus EarthVerse 70.3%
GeoMMAgent + GPT-5.5 GeoMMAgent 68.3%
Seed2.0 Mini EarthVerse 63%
Gemini 3.1 Flash-Lite EarthVerse 59.7% $43
MiniMax-M2.7 EarthVerse 44.7%
GPT-5.6 Luna Codex 35.7% $381
Qwen3-4B 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.

Google DeepMind DeepSeek Anthropic OpenAI Cost frontier One model at each effort level
0% 20% 40% 60% 80% 100% $10 $100 $1,000 $10,000 Cost of the full run, log scale Claude Fable 5 GPT-5.6 Terra DeepSeek-V4-Flash Gemini 3.1 Flash-Lite 0% 20% 40% 60% 80% 100% $10 $100 $1,000 $10,000 Cost of the full run, log scale Claude Fable 5 DeepSeek-V4-Flash

Source: Official EarthVerse leaderboard

ModelEffortHarnessScoreCost of the full runOutput 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%
DeepSeek-V4-Flash EarthVerse 73.5% $44
Tencent Hy3 EarthVerse 72.8%
Claude Haiku 4.5 EarthVerse 72.2% $966
Qwen3.7-Plus EarthVerse 70.3%
Seed2.0 Mini EarthVerse 63%
Gemini 3.1 Flash-Lite EarthVerse 59.7% $43
MiniMax-M2.7 EarthVerse 44.7%
GPT-5.6 Luna Codex 35.7% $381
Qwen3-4B 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.

Anthropic
Own leaderboard Not tested since Aug 2026

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.

Plot with other benchmarks
On the benchmark’s own leaderboard Best so far
0% 20% 40% 60% 80% 100% 2025 2026 Near the ceiling (85%) Claude Fable 5 GPT-5.5 Claude Haiku 4.5 Qwen3-4B 0% 20% 40% 60% 80% 100% 2025 2026 Near the ceiling (85%) Claude Fable 5 GPT-5.5

Source: Official EarthVerse leaderboard

On the benchmark’s own leaderboard Best so far
0% 20% 40% 60% 80% 100% 2025 2026 Near the ceiling (85%) GLM-5.2 Claude Sonnet 4.6 Claude Haiku 4.5 Qwen3-4B 0% 20% 40% 60% 80% 100% 2025 2026 Near the ceiling (85%) GLM-5.2 Claude Haiku 4.5

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

All 203 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.

How we check