Benchmark Almanac

GeoGuesser Bench

Qwen3.5-35B-A3B 22.8 leads since November 2026
Not in model cards

Agents explore offline Street View panoramas to report coordinates, name places, and find objects. Coordinate guesses earn distance points. Other answers require an exact match.

The authors rank 23 models on 282 tasks. The 42 World Cup rounds include expert guesses under different time limits and map controls.

Released
November 2026
Kind
Spatial reasoning and counting
Level
Everyday
Scored
Checked automatically
Answers
Agent task
Tasks
Kept private
Measures
Capability
Language
English
Interaction
Agentic
Environment
A simulator
Refresh
One fixed set
Run by
A host
In views
Agents, Vision, Robotics
Strongest model tested
Claude Fable 5 #12 on the capability index
Last model added
June 2026
Scores refreshed
November 20th, 2026
23
Models scored on its own leaderboard

Qwen3.5-35B-A3B leads with 22.8 among models released by June 2026

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 June 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 Qwen3.5-35B-A3B 22.8
  2. 2 Qwen3.5-122B-A10B 22.0
  3. 3 Grok 4.3 20.0
  4. 4 GPT-5.2 17.9
  5. 5 Qwen3.5 397B-A17B 17.5
  6. 6 Kimi K2.6 14.4
  7. 7 Kimi K2.5 14.1
  8. 8 Gemini 3.1 Pro 14.0
  9. 9 Gemini 3.5 Flash 13.9
  10. 10 Claude Sonnet 4.6 12.6
  11. 11 Gemini 3 Flash 12.4
  12. 12 GPT-5.5 Pro 12.4
  13. 13 Qwen 3.6 Plus 12.3
  14. 14 MiniMax-M3 11.9
  15. 15 Claude Opus 4.6 9.63
  16. 16 GLM-5V-Turbo 9.38
  17. 17 GPT-5.4 8.00
  18. 18 GPT-5.4 Pro 7.75
  19. 19 Qwen3.7-Plus 7.38
  20. 20 Claude Fable 5 5.50
  21. 21 GPT-5.5 3.88
  22. 22 Claude Opus 4.7 3.38
  23. 23 Claude Opus 4.8 3.13

Source: Official GeoGuesser Bench leaderboard

#ModelScoreSettingDateSource
1 Qwen3.5-35B-A3B 22.8 Feb 2026 Official leaderboard
2 Qwen3.5-122B-A10B 22.0 Feb 2026 Official leaderboard
3 Grok 4.3 20.0 Apr 2026 Official leaderboard
4 GPT-5.2 17.9 Dec 2025 Official leaderboard
5 Qwen3.5 397B-A17B 17.5 Feb 2026 Official leaderboard
6 Kimi K2.6 14.4 Apr 2026 Official leaderboard
7 Kimi K2.5 14.1 Jan 2026 Official leaderboard
8 Gemini 3.1 Pro 14.0 Feb 2026 Official leaderboard
9 Gemini 3.5 Flash 13.9 May 2026 Official leaderboard
10 Claude Sonnet 4.6 12.6 Feb 2026 Official leaderboard
11 Gemini 3 Flash 12.4 Dec 2025 Official leaderboard
12 GPT-5.5 Pro 12.4 Apr 2026 Official leaderboard
13 Qwen 3.6 Plus 12.3 Mar 2026 Official leaderboard
14 MiniMax-M3 11.9 Jun 2026 Official leaderboard
15 Claude Opus 4.6 9.63 Feb 2026 Official leaderboard
16 GLM-5V-Turbo 9.38 Apr 2026 Official leaderboard
17 GPT-5.4 8.00 Mar 2026 Official leaderboard
18 GPT-5.4 Pro 7.75 Mar 2026 Official leaderboard
19 Qwen3.7-Plus 7.38 Jun 2026 Official leaderboard
20 Claude Fable 5 5.50 Jun 2026 Official leaderboard
21 GPT-5.5 3.88 Apr 2026 Official leaderboard
22 Claude Opus 4.7 3.38 Apr 2026 Official leaderboard
23 Claude Opus 4.8 3.13 May 2026 Official leaderboard

Rankings by part

We read 4 parts from the benchmark’s own leaderboard in 3 groups: precise localization, place naming and navigation challenge. Each ranks models on one part of the test. Claude Opus 4.8 leads 2 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.

Competition

23 models we track have a score here.

  1. 1 Claude Opus 4.8 ↑22 4352.4
  2. 2 Claude Opus 4.7 ↑20 4306.4
  3. 3 Qwen3.7-Plus ↑16 3964.4
  4. 4 GPT-5.4 ↑13 3923.1
  5. 5 GPT-5.5 ↑16 3815.1
  6. 6 GLM-5V-Turbo ↑10 3138.2
  7. 7 MiniMax-M3 ↑7 3061.5
  8. 8 Qwen 3.6 Plus ↑5 3045.5
  9. 9 Claude Opus 4.6 ↑6 3023.9
  10. 10 Gemini 3 Flash ↑1 2851.0

Number one over time

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

Qwen
Own leaderboard 1 leader

Source: Official GeoGuesser Bench 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
0.00 5.00 10.0 15.0 20.0 25.0 2026 0.00 5.00 10.0 15.0 20.0 25.0 2026

Source: Official GeoGuesser 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 Vision

All 636 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

All 9 checks pass. Every benchmark

Catalog record 2 pass
  • Pass Task access The tasks are private.
  • Pass Confidence Checked against a primary source.
Scores 3 pass 2 notes
  • Pass Independent runs Scores from the official leaderboard (23 entries).
  • Pass Models scored 23 models we track.
  • Pass Newest score The newest is from June 2026, for Claude Fable 5.
  • Note Frontier models None of the top 10 has a score. The strongest tested is Claude Fable 5 (#12).
  • Note Headroom Scored as a plain number, with no fixed ceiling.
Leaderboard import 3 pass
  • Pass Importer We read the board on huanzhimao.com: 23 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.

How we check