Benchmark Almanac

GeoGuessBench (Nutlope)

Claude Opus 5.5 23412.0 leads since October 2026
Not in model cards

Models place a pin after seeing one street photo. Each game has five rounds, and closer guesses earn more points.

The maker scores 12 models on the same 210 photos from Panoramax and KartaView. The leaderboard reports average points per five-round game.

Made by
Nutlope
Released
October 2026
Kind
Spatial reasoning and counting
Level
Everyday
Scored
Checked automatically
Answers
Prediction
Tasks
All public
Measures
Capability
Language
English
In and out
Images → Maps and geodata
Interaction
Single turn
Items from
Real work and use
Refresh
One fixed set
Run by
Anyone
In views
Vision
Strongest model tested
GPT-6 Astra #4 on the capability index
Last model added
September 2026
Scores refreshed
November 20th, 2026
12
Models scored on its own leaderboard

Claude Opus 5.5 leads with 23412.0 among models released by September 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. Every run is in the next chart.

The newest model with a score here came out in September 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 Opus 5.5 Medium effort 23412.0
  2. 2 Claude Fable 5.1 Medium effort 23307.0
  3. 3 Muse Glimmer Medium effort 22407.0
  4. 4 GPT-6 Astra Medium effort 21562.0
  5. 5 GPT-6.1 Sol Medium effort 21194.0
  6. 6 GLM-5.3-Flash Medium effort 21183.0
  7. 7 Kimi K3 Medium effort 20431.0
  8. 8 GPT-6 Luna Medium effort 19560.0
  9. 9 MiniMax-M3 Medium effort 18451.0
  10. 10 Qwen3.8 Flash Medium effort 18169.0
  11. 11 DeepSeek V4.1 Flash Effort 50 of 100 17565.0
  12. 12 Qwen3.5-9B Medium effort 13441.0

Source: Official GeoGuessBench (Nutlope) leaderboard

#ModelScore ±SettingDateSource
1 Claude Opus 5.5 23412.0 23005 to 23755 Medium effort Sep 2026 Official leaderboard
2 Claude Fable 5.1 23307.0 22833 to 23723 Medium effort Sep 2026 Official leaderboard
3 Muse Glimmer 22407.0 21884 to 22903 Medium effort Aug 2026 Official leaderboard
4 GPT-6 Astra 21562.0 20750 to 22314 Medium effort Sep 2026 Official leaderboard
5 GPT-6.1 Sol 21194.0 20342 to 21994 Medium effort Sep 2026 Official leaderboard
6 GLM-5.3-Flash 21183.0 20313 to 21985 Medium effort Aug 2026 Official leaderboard
7 Kimi K3 20431.0 19559 to 21260 Medium effort Jul 2026 Official leaderboard
8 GPT-6 Luna 19560.0 18635 to 20390 Medium effort Sep 2026 Official leaderboard
9 MiniMax-M3 18451.0 ±987.5 Medium effort Jun 2026 Official leaderboard
10 Qwen3.8 Flash 18169.0 ±959 Medium effort Aug 2026 Official leaderboard
11 DeepSeek V4.1 Flash 17565.0 16528 to 18555 Effort 50 of 100 Sep 2026 Official leaderboard
12 Qwen3.5-9B 13441.0 ±1245.5 Medium effort Feb 2026 Official leaderboard

Rankings by part

We read 7 parts from the benchmark’s own leaderboard in 2 groups: location type and region. Each ranks models on one part of the test. Claude Fable 5.1 leads 4 of them, and Claude Opus 5.5 leads the other 3. Arrows show how far a model moved from its overall rank among the models we track.

city

12 models we track have a score here.

  1. 1 Claude Opus 5.5 24137.0
  2. 2 Claude Fable 5.1 23903.0
  3. 3 Muse Glimmer 23200.0
  4. 4 GPT-6 Astra 22428.0
  5. 5 GPT-6.1 Sol 22166.0
  6. 6 Kimi K3 ↑1 21339.0
  7. 7 GLM-5.3-Flash ↓1 21325.0
  8. 8 GPT-6 Luna 20458.0
  9. 9 Qwen3.8 Flash ↑1 18798.0
  10. 10 MiniMax-M3 ↓1 18782.0

The best score, 23412.0, costs $0.06 a task

Runs on the cost frontier score higher than any run that's cheaper. Muse Glimmer at Medium effort comes within 1005 of the best for $0.005 a task, 13 times cheaper. The costs are the leaderboard's own estimates, so compare them only within this benchmark.

Meta Qwen Anthropic OpenAI Cost frontier One model at each effort level
0.00 5000.0 10000.0 15000.0 20000.0 25000.0 $0.001 $0.01 $0.10 $1.00 Cost per task, log scale Claude Opus 5.5, Medium effort Muse Glimmer, Medium effort GPT-6 Luna, Medium effort Qwen3.8 Flash, Medium effort 0.00 5000.0 10000.0 15000.0 20000.0 25000.0 $0.001 $0.01 $0.10 $1.00 Cost per task, log scale Claude Opus 5.5, Medium effort GPT-6 Luna, Medium effort

Source: Official GeoGuessBench (Nutlope) leaderboard

ModelEffortHarnessScoreCost per taskOutput tokens
Claude Opus 5.5 Medium effort 23412.0 $0.06 2,015
Claude Fable 5.1 Medium effort 23307.0 $0.11 1,095
Muse Glimmer Medium effort 22407.0 $0.005 2,045
GPT-6 Astra Medium effort 21562.0 $0.22 3,465
GPT-6.1 Sol Medium effort 21194.0 $0.04 3,210
GLM-5.3-Flash Medium effort 21183.0 $0.009 15,770
Kimi K3 Medium effort 20431.0 $0.17 11,180
GPT-6 Luna Medium effort 19560.0 $0.004 6,065
MiniMax-M3 Medium effort 18451.0 $0.005 2,730
Qwen3.8 Flash Medium effort 18169.0 $0.003 3,925
DeepSeek V4.1 Flash Effort 50 of 100 17565.0 $0.07 56,875
Qwen3.5-9B Medium effort 13441.0 $0.01 44,970

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 Oct 2026
Claude Opus 5.5 23412.0

Source: Official GeoGuessBench (Nutlope) 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
12500.0 15000.0 17500.0 20000.0 22500.0 2026 12500.0 15000.0 17500.0 20000.0 22500.0 2026

Source: Official GeoGuessBench (Nutlope) 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 10 checks pass. Every benchmark

Catalog record 2 pass
  • Pass Task access All tasks are public.
  • Pass Confidence Checked against a primary source.
Scores 4 pass 1 note
  • Pass Independent runs Scores from the official leaderboard (12 entries).
  • Pass Models scored 12 models we track.
  • Pass Newest score The newest is from September 2026, for GPT-6.1 Sol.
  • Pass Frontier models 4 of the top 10 have a score, led by GPT-6 Astra (#4).
  • Note Headroom Scored as a plain number, with no fixed ceiling.
Leaderboard import 3 pass
  • Pass Importer We read the board on geoguessbench.com: 12 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