Benchmark Almanac

SHROOM-Visions 2026

Gemma 4 + Mistral-Small + Qwen3-VL + Qwen3.6 0.46 tops its leaderboard
Not in model cards

Models inspect an image and a written response. They mark unsupported characters, estimate the probability of an error and classify each error in four languages.

The organizers report 27 teams and 623 submissions. SKstars ran two Qwen2.5-VL models. The final tables score span overlap and character-level correlation.

Released
April 2026
Kind
Faithful to sources
Level
Everyday
Scored
Checked automatically
Answers
Prediction
Tasks
Public, plus a private test set
Measures
Capability
In and out
Text, Images → Text
Interaction
Single turn
Items from
Generated
Refresh
One fixed set
Run by
A host
In views
Vision
17
Entries on its own leaderboard, 0 models we track

Gemma 4 + Mistral-Small + Qwen3-VL + Qwen3.6 tops the leaderboard with 0.46 No scores yet

Best score per entry from the benchmark's own leaderboard, read on November 20th, 2026. The organizers report no overall score, so this is the plain mean of labeled correlation across four languages. Where a model ran on its own with several settings, we keep its best. Agents other companies built around a model get a pale shade of that model's lab color. Gray marks blends and entries not in our model list, such as fine-tunes and older models. None of these count toward leaders. Agents built around a model, such as TÜRKSAT + Gemini 3.5 Flash, are their own entries. We don't count them as the model's score. This view shows only models we track, each at its best.

  1. 1 Gemma 4 + Mistral-Small + Qwen3-VL + Qwen3.6 0.46
  2. 2 TÜRKSAT + Gemini 3.5 Flash 0.45
  3. 3 champ + Qwen3.5-27B 0.42
  4. 4 dynamos (Qwen2.5-VL-7B + Qwen2.5-VL-32B) 0.42
  5. 5 SmurfCat + Qwen3-VL-8B-Instruct 0.40
  6. 6 eye-be-am (Qwen3.5-4B + Qwen3.5 397B-A17B) 0.39
  7. 7 caml + Qwen3.5-4B 0.38
  8. 8 HalluVision + Qwen2.5-VL-3B 0.36
  9. 9 Bit-by-bit (Qwen2.5-VL-7B + InternVL3.5-8B) 0.36
  10. 10 Sckwoky + Gemma 4 12B 0.34
  11. 11 XLM-RoBERTa-large + Qwen2-VL-2B-Instruct 0.32
  12. 12 SpanCalib-VLM 0.29
  13. 13 MycoGate-VL 0.27
  14. 14 CPS Lab + Qwen3-VL-8B-Instruct 0.27
  15. 15 mDeBERTa + SigLIP 0.25
  16. 16 XLM-RoBERTa + SigLIP 0.24
  17. 17 WINNAR749 + Qwen2.5-VL-3B 0.19

Source: Official SHROOM-Visions 2026 leaderboard

#EntryScoreSettingSubmitted byDateSource
1 Gemma 4 + Mistral-Small + Qwen3-VL + Qwen3.6 0.46 vroom-vroom vroom-vroom Official leaderboard
2 TÜRKSAT + Gemini 3.5 Flash 0.45 TÜRKSAT May 2026 Official leaderboard
3 champ + Qwen3.5-27B 0.42 champ Feb 2026 Official leaderboard
4 dynamos (Qwen2.5-VL-7B + Qwen2.5-VL-32B) 0.42 dynamos Jan 2025 Official leaderboard
5 SmurfCat + Qwen3-VL-8B-Instruct 0.40 SmurfCat Oct 2025 Official leaderboard
6 eye-be-am (Qwen3.5-4B + Qwen3.5 397B-A17B) 0.39 eye-be-am Feb 2026 Official leaderboard
7 caml + Qwen3.5-4B 0.38 caml Feb 2026 Official leaderboard
8 HalluVision + Qwen2.5-VL-3B 0.36 HalluVision Jan 2025 Official leaderboard
9 Bit-by-bit (Qwen2.5-VL-7B + InternVL3.5-8B) 0.36 Bit-by-bit Jan 2025 Official leaderboard
10 Sckwoky + Gemma 4 12B 0.34 Sckwoky Jun 2026 Official leaderboard
11 XLM-RoBERTa-large + Qwen2-VL-2B-Instruct 0.32 USP USP Official leaderboard
12 SpanCalib-VLM 0.29 SIT SIT Official leaderboard
13 MycoGate-VL 0.27 L1cache L1cache Official leaderboard
14 CPS Lab + Qwen3-VL-8B-Instruct 0.27 CPS Lab Oct 2025 Official leaderboard
15 mDeBERTa + SigLIP 0.25 Scalar_Nitk Scalar_Nitk Official leaderboard
16 XLM-RoBERTa + SigLIP 0.24 DSR DSR Official leaderboard
17 WINNAR749 + Qwen2.5-VL-3B 0.19 WINNAR749 Jan 2025 Official leaderboard
#ModelScoreSettingSubmitted byDateSource

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 Factuality

All 164 in Factuality
  • SimpleVQA In use Models answer 2,025 short factual questions about images across 9 tasks and 9 topics. An LLM judge grades answers against fixed references. Qwen3.7-Plus 81.7%
  • HallusionBench In use Models answer 1,129 yes or no questions about 346 images, built in pairs to expose language hallucination and visual illusion. Scored by accuracy. Keye-VL-2.0-30B-A3B 73.5%
  • SimpleQA Verified In use 1,000 human-verified, de-duplicated short-answer factual questions measuring parametric knowledge. Gemini 3.1 Pro 77.5%
  • AA-Omniscience v1.1 In use Models answer 3,000 knowledge questions in six domains without tools and may abstain. The index adds correct answers and subtracts wrong ones. Inkling 2 Small 16.1
  • GaRAGe In use Models write long-form RAG answers to 2,366 questions with human-annotated grounding passages. Scoring checks whether answers use the relevant passages. Amazon Nova Pro 60.7%
  • Kimi Hallu Bench In use Models answer hard prompts drawn from practical work tasks. Fact-checking agents score the share of claims that are hallucinated. GPT-6 Astra 4.20%

Data audit

All 8 checks pass. Every benchmark

Catalog record 2 pass
  • Pass Task access Public tasks, plus a private test set.
  • Pass Confidence Checked against a primary source.
Scores 2 pass 3 notes
  • Pass Independent runs Scores from the official leaderboard (17 entries).
  • Pass Models scored 17 entries on its leaderboard, none a model we track.
  • Note Newest score No scores yet.
  • Note Frontier models No models scored yet.
  • Note Headroom Scored as a plain number, with no fixed ceiling.
Leaderboard import 3 pass
  • Pass Importer We read the board on helsinki-nlp.github.io: 11 rows for models we track and 6 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