Benchmark Almanac

GeoBiaset

Qwen3VL-30B-A3B 72.9% tops its leaderboard GLM-4.1V-9B-Thinking 69.2% leads the models we track
Not in model cards

Models predict photo coordinates before and after researchers insert a person into the scene. The coordinates stay fixed across all edits of each background.

The authors compare seven systems across 329 backgrounds and 9,212 edits. They report location accuracy at several distance thresholds.

Kind
Bias, fairness and politics
Level
University
Scored
Checked automatically
Answers
Short answer
Tasks
All public
Measures
Propensity
Behavior
Bias
Language
English
In and out
Text, Images → Maps and geodata
Interaction
Single turn
Refresh
One fixed set
Run by
Anyone
In views
Vision
Strongest model tested
Gemma 3 27B #165 on the capability index
Last model added
October 2025
Scores refreshed
November 20th, 2026
10
Entries on its own leaderboard, 5 models we track

Qwen3VL-30B-A3B tops the leaderboard with 72.9% GLM-4.1V-9B-Thinking leads with 69.2% among models released by October 2025

Best score per entry from the benchmark's own leaderboard, read on November 20th, 2026. The headline is continent-level accuracy on the 9,212 edited images. The original backgrounds provide the comparison. 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. The best model we track is GLM-4.1V-9B-Thinking, at 69.2%. This view shows only models we track, each at its best.

The newest model with a score here came out in October 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 Qwen3VL-30B-A3B 72.9%
  2. 2 GLM-4.1V-9B-Thinking 69.2%
  3. 3 Gemma 3 27B 67.8%
  4. 4 Qwen3-VL-8B-Instruct 65.3%
  5. 5 InternVL3.5-38B 55.4%
  6. 6 GeoCLIP 55.3%
  7. 7 DeepSeek-VL 51.5%
  8. 8 G3 47.8%
  9. 9 InternVL3.5-8B 43.2%
  10. 10 Img2Loc 37.5%

Source: Official GeoBiaset leaderboard

#EntryScoreSettingDateSource
1 Qwen3VL-30B-A3B 72.9% Official leaderboard
2 GLM-4.1V-9B-Thinking 69.2% Jul 2025 Official leaderboard
3 Gemma 3 27B 67.8% Mar 2025 Official leaderboard
4 Qwen3-VL-8B-Instruct 65.3% Oct 2025 Official leaderboard
5 InternVL3.5-38B 55.4% Aug 2025 Official leaderboard
6 GeoCLIP 55.3% Official leaderboard
7 DeepSeek-VL 51.5% Official leaderboard
8 G3 47.8% Official leaderboard
9 InternVL3.5-8B 43.2% Aug 2025 Official leaderboard
10 Img2Loc 37.5% Official leaderboard
#ModelScoreSettingDateSource
1 GLM-4.1V-9B-Thinking 69.2% Jul 2025 Official leaderboard
2 Gemma 3 27B 67.8% Mar 2025 Official leaderboard
3 Qwen3-VL-8B-Instruct 65.3% Oct 2025 Official leaderboard
4 InternVL3.5-38B 55.4% Aug 2025 Official leaderboard
5 InternVL3.5-8B 43.2% Aug 2025 Official leaderboard

Rankings by part

We read 4 parts from the benchmark’s own leaderboard, each ranking models on one part of the test, such as Original accuracy, POPI-A and POPI-H. Gemma 3 27B 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.

Original accuracy

5 models we track have a score here.

  1. 1 GLM-4.1V-9B-Thinking 82.8%
  2. 2 Gemma 3 27B 80.6%
  3. 3 Qwen3-VL-8B-Instruct 79.3%
  4. 4 InternVL3.5-38B 71.1%
  5. 5 InternVL3.5-8B 58%

2 models have held number one on the official leaderboard since 2023

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

Google DeepMind Z.ai
Own leaderboard Not tested since Oct 2025

Source: Official GeoBiaset 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 Near the ceiling (85%) GLM-4.1V-9B-Thinking Gemma 3 27B 0% 20% 40% 60% 80% 100% 2025 Near the ceiling (85%) GLM-4.1V-9B-Thinking

Source: Official GeoBiaset 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 Safety and alignment

All 636 in Safety and alignment
  • FORTRESS In use Paired adversarial and benign national-security misuse prompts. Rubric-graded risk score on adversarial and over-refusal score on benign prompts. Claude Opus 5.5 5.78%
  • StrongREJECT In use Forbidden-prompt jailbreak benchmark with a rubric grader that scores how useful a non-refusal actually is. Inkling 98.6%
  • Gray Swan IPI Arena benchmark In use Indirect prompt-injection attacks from public red-teaming competitions replayed on tool, coding and browser agents. Attack success rate at k attempts. Gemini 4 Carbon 0%
  • Covert misbehavior (Thinking Machines) In use Agentic tasks that check whether a model misbehaves and conceals it. Scored by the rate of covert misbehavior. Inkling 2 Mini 8.30%
  • Destructive actions (Thinking Machines) In use Agentic tasks that check whether a model takes destructive actions it was not authorized to take. Scored by the rate of destructive actions. Inkling 2 Small 4.70%
  • Misbehavior (Thinking Machines) In use Agentic tasks that check whether a model fabricates results, ignores instructions or hides what it did when a task is hard. Scored by the rate of misbehavior. Inkling 2 Mini 9.80%

Data audit

9 of 10 checks pass, 1 warning. Look first at the Newest score check. The newest came 13 months ago, in October 2025, for Qwen3-VL-8B-Instruct. 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 (10 entries).
  • Pass Models scored 5 models we track, plus 5 other entries on its leaderboard.
  • Warning Newest score The newest came 13 months ago, in October 2025, for Qwen3-VL-8B-Instruct.
  • Note Frontier models None of the top 10 has a score. The strongest tested is Gemma 3 27B (#165).
  • Pass Headroom The best is 69.2%, under 85%, so it still separates models.
Leaderboard import 3 pass
  • Pass Importer We read the board on miccunifi.github.io: 5 rows for models we track and 5 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