Benchmark Almanac

CREATE

Gemma 4 12B 42.2% leads since October 2026
Not in model cards

Models find a word that connects three photographs. A judge grades each answer against the intended connection.

The authors publish five runs each for six vision language models on 155 puzzles. Twenty-five people answered a subset of 80 puzzles.

Released
October 2026
Kind
General visual questions
Level
Everyday
Scored
Graded by a model
Answers
Short answer
Tasks
All public
Measures
Capability
Language
English
In and out
Text, Images → Text
Interaction
Single turn
Items from
Generated
Refresh
One fixed set
Run by
Anyone
In views
Vision
Strongest model tested
Qwen 3.8 27B #59 on the capability index
Last model added
August 2026
Scores refreshed
November 20th, 2026
6
Entries on its own leaderboard, 4 models we track

Gemma 4 12B leads with 42.2% among models released by August 2026 Gemma 4 12B leads with 42.2% among models released by August 2026

Best score per entry 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. 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 August 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 Gemma 4 12B text_only 42.2%
  2. 2 Qwen 3.8 27B text_only 39.2%
  3. 3 RedHatAI/Llama-4-Scout-17B-16E-Instruct-quantized.w4a16 text_only 34.1%
  4. 4 Qwen3.5-9B text_only 29.2%
  5. 5 Molmo2-8B text_only 26.3%
  6. 6 OpenGVLab/InternVL3_5-8B-HF text_only 20.6%

Source: Official CREATE leaderboard

#EntryScoreSettingDateSource
1 Gemma 4 12B 42.2% text_only Jun 2026 Official leaderboard
2 Qwen 3.8 27B 39.2% text_only Aug 2026 Official leaderboard
3 RedHatAI/Llama-4-Scout-17B-16E-Instruct-quantized.w4a16 34.1% text_only Official leaderboard
4 Qwen3.5-9B 29.2% text_only Feb 2026 Official leaderboard
5 Molmo2-8B 26.3% text_only Dec 2025 Official leaderboard
6 OpenGVLab/InternVL3_5-8B-HF 20.6% text_only Official leaderboard
#ModelScoreSettingDateSource
1 Gemma 4 12B 42.2% text_only Jun 2026 Official leaderboard
2 Qwen 3.8 27B 39.2% text_only Aug 2026 Official leaderboard
3 Qwen3.5-9B 29.2% text_only Feb 2026 Official leaderboard
4 Molmo2-8B 26.3% text_only Dec 2025 Official leaderboard

Scores at each effort level

Models the leaderboard ran at two or more effort levels. We don’t have enough cost data here to say what the extra effort costs.

Lowest effort to highest, light to dark
Gemma 4 12B
stepwise zeroshot text_only
42.2%
Qwen 3.8 27B
stepwise detailed two_shot text_only
39.2%
Qwen3.5-9B
zeroshot two_shot text_only
29.2%
Molmo2-8B
two_shot text_only
26.3%

Source: Official CREATE leaderboard

ModelEffortHarnessScoreCost per taskOutput tokens
Gemma 4 12B text_only 42.2%
Qwen 3.8 27B text_only 39.2%
Qwen3.5-9B text_only 29.2%
Molmo2-8B text_only 26.3%
Qwen 3.8 27B two_shot 12.7%
Gemma 4 12B zeroshot 9.68%
Qwen 3.8 27B detailed 5.55%
Qwen3.5-9B two_shot 4.26%
Gemma 4 12B two_shot 4.26%
Qwen 3.8 27B zeroshot 3.35%
Molmo2-8B zeroshot 2.58%
Molmo2-8B stepwise 2.58%
Gemma 4 12B detailed 2.58%
Gemma 4 12B stepwise 2.06%
Molmo2-8B detailed 1.94%
Qwen 3.8 27B stepwise 1.42%
Qwen3.5-9B detailed 1.03%
Qwen3.5-9B stepwise 1.03%
Qwen3.5-9B zeroshot 0.65%
Molmo2-8B two_shot 0.52%

Scores at each effort level

Models the leaderboard ran at two or more effort levels. We don’t have enough cost data here to say what the extra effort costs.

Lowest effort to highest, light to dark
Gemma 4 12B
stepwise zeroshot text_only
42.2%
Qwen 3.8 27B
stepwise detailed two_shot text_only
39.2%
Qwen3.5-9B
zeroshot two_shot text_only
29.2%
Molmo2-8B
two_shot text_only
26.3%

Source: Official CREATE leaderboard

ModelEffortHarnessScoreCost per taskOutput tokens
Gemma 4 12B text_only 42.2%
Qwen 3.8 27B text_only 39.2%
Qwen3.5-9B text_only 29.2%
Molmo2-8B text_only 26.3%
Qwen 3.8 27B two_shot 12.7%
Gemma 4 12B zeroshot 9.68%
Qwen 3.8 27B detailed 5.55%
Qwen3.5-9B two_shot 4.26%
Gemma 4 12B two_shot 4.26%
Qwen 3.8 27B zeroshot 3.35%
Molmo2-8B zeroshot 2.58%
Molmo2-8B stepwise 2.58%
Gemma 4 12B detailed 2.58%
Gemma 4 12B stepwise 2.06%
Molmo2-8B detailed 1.94%
Qwen 3.8 27B stepwise 1.42%
Qwen3.5-9B detailed 1.03%
Qwen3.5-9B stepwise 1.03%
Qwen3.5-9B zeroshot 0.65%
Molmo2-8B two_shot 0.52%

Number one over time

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

Google DeepMind
Own leaderboard Not tested since Oct 2026
Gemma 4 12B 42.2%

Source: Official CREATE 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% 2026 Near the ceiling (85%) Gemma 4 12B Qwen3.5-9B 0% 20% 40% 60% 80% 100% 2026 Near the ceiling (85%) Gemma 4 12B Qwen3.5-9B

Source: Official CREATE 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 (6 entries).
  • Pass Models scored 4 models we track, plus 2 other entries on its leaderboard.
  • Pass Newest score The newest is from August 2026, for Qwen 3.8 27B.
  • Note Frontier models None of the top 10 has a score. The strongest tested is Qwen 3.8 27B (#59).
  • Pass Headroom The best is 42.2%, under 85%, so it still separates models.
Leaderboard import 3 pass
  • Pass Importer We read the board on github.com: 20 rows for models we track and 2 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