Benchmark Almanac

WeatherQA (severe weather)

Qwen2.5-VL-7B 62.7% leads since January 2025
Not in model cards

Models read weather maps to identify a threatened area and classify its severe weather potential. The second score includes only cases with the correct area.

The maintainers list 23 scored entries per task, including newer GPT and Gemini models. The paper samples 600 cases for these tasks.

Released
June 2024
Kind
Science questions
Level
Expert
Scored
Checked automatically
Answers
Multiple choice
Tasks
All public
Measures
Capability
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
GPT-5.2 #51 on the capability index
Last model added
December 2025
Scores refreshed
November 20th, 2026
22
Entries on its own leaderboard, 10 models we track

Qwen2.5-VL-7B leads with 62.7% among models released by December 2025 Qwen2.5-VL-7B leads with 62.7% among models released by December 2025

Best score per entry from the benchmark's own leaderboard, read on November 20th, 2026. The headline averages affected-area accuracy and severe weather concern accuracy conditional on a correct affected area. 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. This view shows only models we track, each at its best.

The newest model with a score here came out in December 2025, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

  1. 1 Qwen2.5-VL-7B Zero-shot, Three Images (rgnlrad, sbcp, shr6), Fine-tuned 62.7%
  2. 2 Qwen3-VL-4B Zero-shot, fine-tuned 60.5%
  3. 3 Gemini 3.0 Zero-shot 34.8%
  4. 4 GPT-5.1 Zero-shot 30.8%
  5. 5 GPT-5.2 Zero-shot 29.1%
  6. 6 Gemini 2.0 Zero-shot 27.9%
  7. 7 Gemini Flash 2.0 Zero-shot 27.8%
  8. 8 Gemini 2.5 Zero-shot 25.4%
  9. 9 Gemini Flash 2.5 Zero-shot 25.4%
  10. 10 Claude 3.5 Sonnet (snapshot unspecified) Zero-shot 25.1%
  11. 11 GPT-4o (snapshot unspecified) Zero-shot 21.9%
  12. 12 Gemini 1.5 (snapshot unspecified) Zero-shot 20.5%
  13. 13 Gemini Pro 1.5 (snapshot unspecified) Zero-shot 17.8%
  14. 14 GPT-4T Zero-shot 14.9%
  15. 15 InternVL2-40B Zero-shot 14.2%
  16. 16 InternVL2-8B Zero-shot 13.6%
  17. 17 Qwen2-VL-7B-Instruct Zero-shot 12.8%
  18. 18 InternVL2-26B Zero-shot 12.8%
  19. 19 GPT-4 Turbo (snapshot unspecified) Zero-shot 12.7%
  20. 20 Claude 3 Opus Zero-shot 11%
  21. 21 Phi-3.5-vision-instruct Zero-shot 10.7%
  22. 22 Qwen2.5-VL-3B Zero-shot 9.70%

Source: Official WeatherQA (severe weather) leaderboard

#EntryScoreSettingDateSource
1 Qwen2.5-VL-7B 62.7% Zero-shot, Three Images (rgnlrad, sbcp, shr6), Fine-tuned Jan 2025 Official leaderboard
2 Qwen3-VL-4B 60.5% Zero-shot, fine-tuned Official leaderboard
3 Gemini 3.0 34.8% Zero-shot Official leaderboard
4 GPT-5.1 30.8% Zero-shot Nov 2025 Official leaderboard
5 GPT-5.2 29.1% Zero-shot Dec 2025 Official leaderboard
6 Gemini 2.0 27.9% Zero-shot Official leaderboard
7 Gemini Flash 2.0 27.8% Zero-shot Official leaderboard
8 Gemini 2.5 25.4% Zero-shot Official leaderboard
9 Gemini Flash 2.5 25.4% Zero-shot Official leaderboard
10 Claude 3.5 Sonnet (snapshot unspecified) 25.1% Zero-shot Official leaderboard
11 GPT-4o (snapshot unspecified) 21.9% Zero-shot Official leaderboard
12 Gemini 1.5 (snapshot unspecified) 20.5% Zero-shot Official leaderboard
13 Gemini Pro 1.5 (snapshot unspecified) 17.8% Zero-shot Official leaderboard
14 GPT-4T 14.9% Zero-shot Official leaderboard
15 InternVL2-40B 14.2% Zero-shot Jul 2024 Official leaderboard
16 InternVL2-8B 13.6% Zero-shot Jul 2024 Official leaderboard
17 Qwen2-VL-7B-Instruct 12.8% Zero-shot Aug 2024 Official leaderboard
18 InternVL2-26B 12.8% Zero-shot Jul 2024 Official leaderboard
19 GPT-4 Turbo (snapshot unspecified) 12.7% Zero-shot Official leaderboard
20 Claude 3 Opus 11% Zero-shot Feb 2024 Official leaderboard
21 Phi-3.5-vision-instruct 10.7% Zero-shot Aug 2024 Official leaderboard
22 Qwen2.5-VL-3B 9.70% Zero-shot Jan 2025 Official leaderboard
#ModelScoreSettingDateSource
1 Qwen2.5-VL-7B 62.7% Zero-shot, Three Images (rgnlrad, sbcp, shr6), Fine-tuned Jan 2025 Official leaderboard
2 GPT-5.1 30.8% Zero-shot Nov 2025 Official leaderboard
3 GPT-5.2 29.1% Zero-shot Dec 2025 Official leaderboard
4 InternVL2-40B 14.2% Zero-shot Jul 2024 Official leaderboard
5 InternVL2-8B 13.6% Zero-shot Jul 2024 Official leaderboard
6 Qwen2-VL-7B-Instruct 12.8% Zero-shot Aug 2024 Official leaderboard
7 InternVL2-26B 12.8% Zero-shot Jul 2024 Official leaderboard
8 Claude 3 Opus 11% Zero-shot Feb 2024 Official leaderboard
9 Phi-3.5-vision-instruct 10.7% Zero-shot Aug 2024 Official leaderboard
10 Qwen2.5-VL-3B 9.70% Zero-shot Jan 2025 Official leaderboard

Rankings by part

We read 2 parts from the benchmark’s own leaderboard: Areas Affected QA and Conditional Severe Weather Concern Classification. Each ranks models on one part of the test. Qwen2.5-VL-7B leads both. Arrows show how far a model moved from its overall rank among the models we track.

Areas Affected QA

10 models we track have a score here.

  1. 1 Qwen2.5-VL-7B 70.3%
  2. 2 GPT-5.2 ↑1 43.5%
  3. 3 GPT-5.1 ↓1 43%
  4. 4 InternVL2-40B 25.3%
  5. 5 InternVL2-8B 24.2%
  6. 6 InternVL2-26B ↑1 23.7%
  7. 7 Qwen2-VL-7B-Instruct ↓1 22.7%
  8. 8 Claude 3 Opus 20.7%
  9. 9 Phi-3.5-vision-instruct 20%
  10. 10 Qwen2.5-VL-3B 16.3%

4 models have held number one on the official leaderboard since its release in June 2024

Lab claims count from the day the card came out, independent runs from the model’s release.

Anthropic Shanghai AI Lab Qwen
Own leaderboard Not tested since Dec 2025

Source: Official WeatherQA (severe weather) 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% 2024 2025 2026 Near the ceiling (85%) Qwen2.5-VL-7B InternVL2-40B Claude 3 Opus 0% 20% 40% 60% 80% 100% 2024 2025 2026 Near the ceiling (85%) Qwen2.5-VL-7B InternVL2-40B

Source: Official WeatherQA (severe weather) 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 Newest score check. The newest came 11 months ago, in December 2025, for GPT-5.2. 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 (22 entries).
  • Pass Models scored 10 models we track, plus 12 other entries on its leaderboard.
  • Warning Newest score The newest came 11 months ago, in December 2025, for GPT-5.2.
  • Note Frontier models None of the top 10 has a score. The strongest tested is GPT-5.2 (#51).
  • Pass Headroom The best is 62.7%, under 85%, so it still separates models.
Leaderboard import 3 pass
  • Pass Importer We read the board on github.com: 15 rows for models we track and 12 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