PerceptionBench (Moonshot)
The official leaderboard hasn’t scored a model released after Kimi K3 (July 2026). Labs report their own numbers for 5 newer models, such as Muse Ember at 80.6%.
Moonshot in-house benchmark of atomic visual perception skills. Scoring details unpublished.
Lab-created in-house eval (Moonshot AI). Not the same as other public benchmarks with this name.
- Made by
- Moonshot AI
- Released
- July 2026
- Kind
- General visual questions
- Level
- Everyday
- Scored
- Checked automatically
- Answers
- Short answer
- Tasks
- Kept private
- Measures
- Capability
- Interaction
- Single turn
- Items from
- Written by experts
- Refresh
- One fixed set
- Run by
- A host
- In views
- Vision
- Strongest model tested
- Claude Fable 5 #12 on the capability index
- Last in a launch post
- November 2026
- Last model added
- November 2026
- Scores refreshed
- November 18th, 2026
GPT-5.6 Sol leads with 59.7% among models released by July 2026
Best score per model from the benchmark's own leaderboard, read on November 18th, 2026. Where a model ran with several settings or harnesses, we keep its best.
Muse Ember leads with 80.6%, claims included
Best score per model from every source, including labs’ own numbers and claims Epoch AI collected from reports and papers. Labs pick their own settings, so a claim can beat an independent run of the same model. Striped bars are claims, and dotted bars are reruns by Epoch AI or another board.
The newest model with a score here came out in July 2026, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:
- GPT-6.2 Astra
- Claude Fable 5.2
-
Gemini 4 Carbon
- GPT-6 Astra
- and 14 more
- 1 GPT-5.6 Sol 59.7%
- 2 Kimi K3 58.5%
- 3 Claude Fable 5 57.2%
- 4
Gemini 3.1 Pro 56.2%
- 5 GPT-5.5 55.8%
- 6
Seed2.1 Pro 55%
- 7
Gemini 3.5 Flash 52%
- 8
Qwen3.7-Plus 51.1%
- 9
Qwen3.5 397B-A17B 47.5%
- 10 Claude Opus 4.8 47.2%
- 11 Kimi K2.6 42.6%
- 12 Grok 4.5 41%
- 13
Gemma 4 31B IT 40.7%
- 14 GLM-5V-Turbo 39.6%
- 15
MiniMax-M3 33.1%
- 16 GLM-4.6V 32.5%
Source: Official PerceptionBench (Moonshot) leaderboard
- 80.6
Muse Ember
- 68.3 Kimi K3.1 Pro
- 64.7
Inkling 2 Small
- 63.5
Qwen 3.8 Max
- 59.7 GPT-5.6 Sol
- 58.5 Kimi K3
- 57.5
Inkling 2 Mini
- 57.2 Claude Fable 5
- 56.2
Gemini 3.1 Pro
- 55.8 GPT-5.5
- 55.0
Seed2.1 Pro
- 52.0
Gemini 3.5 Flash
- 51.1
Qwen3.7-Plus
- 47.5
Qwen3.5 397B-A17B
- 47.2 Claude Opus 4.8
- 42.6 Kimi K2.6
- 41.0 Grok 4.5
- 40.7
Gemma 4 31B IT
- 39.6 GLM-5V-Turbo
- 33.1
MiniMax-M3
- 32.5 GLM-4.6V
- 1
Muse Ember max effort 80.6%
- 2 Kimi K3.1 Pro 68.3%
- 3
Inkling 2 Small 64.7%
- 4
Qwen 3.8 Max 63.5%
- 5 GPT-5.6 Sol 59.7%
- 6 Kimi K3 58.5%
- 7
Inkling 2 Mini 57.5%
- 8 Claude Fable 5 57.2%
- 9
Gemini 3.1 Pro 56.2%
- 10 GPT-5.5 55.8%
- 11
Seed2.1 Pro 55%
- 12
Gemini 3.5 Flash 52%
- 13
Qwen3.7-Plus 51.1%
- 14
Qwen3.5 397B-A17B 47.5%
- 15 Claude Opus 4.8 47.2%
- 16 Kimi K2.6 42.6%
- 17 Grok 4.5 41%
- 18
Gemma 4 31B IT 40.7%
- 19 GLM-5V-Turbo 39.6%
- 20
MiniMax-M3 33.1%
- 21 GLM-4.6V 32.5%
Sources: Labs’ model cards and official PerceptionBench (Moonshot) leaderboard
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | GPT-5.6 Sol | 59.7% | Jul 2026 | Official leaderboard | |
| 2 | Kimi K3 | 58.5% | Jul 2026 | Official leaderboard | |
| 3 | Claude Fable 5 | 57.2% | Jun 2026 | Official leaderboard | |
| 4 | | 56.2% | Feb 2026 | Official leaderboard | |
| 5 | GPT-5.5 | 55.8% | Apr 2026 | Official leaderboard | |
| 6 | | 55% | Jun 2026 | Official leaderboard | |
| 7 | | 52% | May 2026 | Official leaderboard | |
| 8 | | 51.1% | Jun 2026 | Official leaderboard | |
| 9 | | 47.5% | Feb 2026 | Official leaderboard | |
| 10 | Claude Opus 4.8 | 47.2% | May 2026 | Official leaderboard | |
| 11 | Kimi K2.6 | 42.6% | Apr 2026 | Official leaderboard | |
| 12 | Grok 4.5 | 41% | Jul 2026 | Official leaderboard | |
| 13 | | 40.7% | Apr 2026 | Official leaderboard | |
| 14 | GLM-5V-Turbo | 39.6% | Apr 2026 | Official leaderboard | |
| 15 | | 33.1% | Jun 2026 | Official leaderboard | |
| 16 | GLM-4.6V | 32.5% | Dec 2025 | Official leaderboard |
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | 80.6% | max effort | Nov 2026 | Lab model card | |
| 2 | Kimi K3.1 Pro | 68.3% | Oct 2026 | Lab model card | |
| 3 | 64.7% | Nov 2026 | Lab model card | ||
| 4 | 63.5% | Aug 2026 | Lab model card | ||
| 5 | GPT-5.6 Sol | 59.7% | Jul 2026 | Official leaderboard | |
| 6 | Kimi K3 | 58.5% | Jul 2026 | Official leaderboard | |
| 7 | 57.5% | Nov 2026 | Lab model card | ||
| 8 | Claude Fable 5 | 57.2% | Jun 2026 | Official leaderboard | |
| 9 | 56.2% | Feb 2026 | Official leaderboard | ||
| 10 | GPT-5.5 | 55.8% | Apr 2026 | Official leaderboard | |
| 11 | 55% | Jun 2026 | Official leaderboard | ||
| 12 | 52% | May 2026 | Official leaderboard | ||
| 13 | 51.1% | Jun 2026 | Official leaderboard | ||
| 14 | 47.5% | Feb 2026 | Official leaderboard | ||
| 15 | Claude Opus 4.8 | 47.2% | May 2026 | Official leaderboard | |
| 16 | Kimi K2.6 | 42.6% | Apr 2026 | Official leaderboard | |
| 17 | Grok 4.5 | 41% | Jul 2026 | Official leaderboard | |
| 18 | 40.7% | Apr 2026 | Official leaderboard | ||
| 19 | GLM-5V-Turbo | 39.6% | Apr 2026 | Official leaderboard | |
| 20 | 33.1% | Jun 2026 | Official leaderboard | ||
| 21 | GLM-4.6V | 32.5% | Dec 2025 | Official leaderboard |
Labs report scores for 6 models in their cards. Switch the scores above to rank them together.
Rankings by part
We read 9 parts from the benchmark’s own leaderboard, each ranking models on one part of the test, such as Depth, OCR and Visual relation. GPT-5.6 Sol leads 5 of them, and 3 other models lead the rest. Arrows show how far a model moved from its overall rank among the models we track.
Depth
16 models we track have a score here.
- 1 GPT-5.6 Sol 55.5%
- 2 Kimi K3 52.4%
- 3 Claude Fable 5 51.5%
- 4
Gemini 3.1 Pro 50%
- 5
Gemini 3.5 Flash ↑2 49.7%
- 6 GPT-5.5 ↓1 48.8%
- 7
Qwen3.7-Plus ↑1 48.5%
- 8
Qwen3.5 397B-A17B ↑1 44.6%
- 9
Seed2.1 Pro ↓3 43.6%
- 10 Kimi K2.6 ↑1 42.4%
OCR
16 models we track have a score here.
- 1
Seed2.1 Pro ↑5 66.7%
- 2
Gemini 3.1 Pro ↑2 64.3%
- 3 Claude Fable 5 64.3%
- 4 Kimi K3 ↓2 61.2%
- 5
Gemini 3.5 Flash ↑2 59.6%
- 6 GPT-5.5 ↓1 56.5%
- 7 GPT-5.6 Sol ↓6 54.9%
- 8
Qwen3.7-Plus 54.5%
- 9 Claude Opus 4.8 ↑1 54.1%
- 10
Qwen3.5 397B-A17B ↓1 52.9%
Visual relation
16 models we track have a score here.
- 1 GPT-5.6 Sol 69.7%
- 2 Kimi K3 68.2%
- 3 GPT-5.5 ↑2 61.9%
- 4
Qwen3.7-Plus ↑4 59.1%
- 5
Gemini 3.1 Pro ↓1 58.8%
- 6 Claude Fable 5 ↓3 58.5%
- 7
Seed2.1 Pro ↓1 57.6%
- 8
Qwen3.5 397B-A17B ↑1 55.2%
- 9
Gemini 3.5 Flash ↓2 53.6%
- 10 Claude Opus 4.8 51.4%
Attribute
16 models we track have a score here.
- 1 GPT-5.6 Sol 62.1%
- 2
Gemini 3.1 Pro ↑2 61.8%
- 3 GPT-5.5 ↑2 60.9%
- 4 Claude Fable 5 ↓1 60.9%
- 5 Kimi K3 ↓3 59.4%
- 6
Seed2.1 Pro 58.2%
- 7
Qwen3.7-Plus ↑1 55.8%
- 8
Gemini 3.5 Flash ↓1 54.5%
- 9
Qwen3.5 397B-A17B 53%
- 10 Claude Opus 4.8 49.4%
Localization
16 models we track have a score here.
- 1 GPT-5.6 Sol 76.7%
- 2 Claude Fable 5 ↑1 70.4%
- 3 Kimi K3 ↓1 70.3%
- 4 GPT-5.5 ↑1 65.8%
- 5 Claude Opus 4.8 ↑5 58.8%
- 6
Gemini 3.1 Pro ↓2 52.7%
- 7
Qwen3.7-Plus ↑1 52.7%
- 8
Gemini 3.5 Flash ↓1 50.6%
- 9
Seed2.1 Pro ↓3 50%
- 10
Qwen3.5 397B-A17B ↓1 46.7%
Comparison
16 models we track have a score here.
- 1 GPT-5.6 Sol 67%
- 2 GPT-5.5 ↑3 65.6%
- 3
Gemini 3.1 Pro ↑1 61.7%
- 4
Seed2.1 Pro ↑2 59.5%
- 5 Kimi K3 ↓3 59.1%
- 6 Claude Fable 5 ↓3 56.1%
- 7
Qwen3.7-Plus ↑1 55.9%
- 8
Gemini 3.5 Flash ↓1 54.8%
- 9
Qwen3.5 397B-A17B 49.8%
- 10 Claude Opus 4.8 48.4%
Fine-grained recognition
16 models we track have a score here.
- 1
Seed2.1 Pro ↑5 56.6%
- 2 GPT-5.6 Sol ↓1 55.9%
- 3 Kimi K3 ↓1 55.9%
- 4
Gemini 3.1 Pro 54.8%
- 5
Gemini 3.5 Flash ↑2 51.7%
- 6 Claude Fable 5 ↓3 51.6%
- 7 GPT-5.5 ↓2 47.2%
- 8
Qwen3.7-Plus 46.8%
- 9
Qwen3.5 397B-A17B 44.8%
- 10 Claude Opus 4.8 40.7%
Contextual integration
16 models we track have a score here.
- 1
Gemini 3.1 Pro ↑3 61.2%
- 2
Seed2.1 Pro ↑4 60.4%
- 3 GPT-5.6 Sol ↓2 60%
- 4 Claude Fable 5 ↓1 59.8%
- 5 GPT-5.5 58%
- 6
Gemini 3.5 Flash ↑1 53.3%
- 7 Kimi K3 ↓5 53.3%
- 8
Qwen3.7-Plus 52.2%
- 9
Qwen3.5 397B-A17B 50.2%
- 10
Gemma 4 31B IT ↑3 45.9%
Perception-related hallucination
16 models we track have a score here.
- 1
Gemini 3.5 Flash ↑6 50.6%
- 2
Seed2.1 Pro ↑4 49.8%
- 3 Claude Fable 5 45%
- 4 Grok 4.5 ↑8 44.7%
- 5 Kimi K3 ↓3 41.7%
- 6 Kimi K2.6 ↑5 41%
- 7
Gemini 3.1 Pro ↓3 40.6%
- 8 Claude Opus 4.8 ↑2 38.7%
- 9 GPT-5.5 ↓4 34.7%
- 10
Gemma 4 31B IT ↑3 32.1%
2 models have held number one on the official leaderboard since its release in July 2026 5 models have held number one with claims included since its release in July 2026
Lab claims count from the day the card came out, independent runs from the model’s release. We draw one row each for the benchmark’s own leaderboard and what labs claim in their cards.
Sources: Official PerceptionBench (Moonshot) leaderboard and labs’ model cards
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Sources: Official PerceptionBench (Moonshot) leaderboard and labs’ model cards
4 labs put it in a launch post in the last six months
Launch posts from every lab that has listed it, one tick per post. Tall colored ticks include it, and short gray ones leave it out.
| Date | Launch post | Score | As the post writes it |
|---|---|---|---|
| Nov 20, 2026 | 57.5 | PerceptionBench | |
| Nov 9, 2026 | 80.6 | PerceptionBench with tools, Muse Ember max reasoning | |
| Nov 5, 2026 | 64.7 | PerceptionBench | |
| Oct 24, 2026 | Kimi K3.1 Pro | 68.3 | PerceptionBench |
| Aug 3, 2026 | 63.5 | PerceptionBench | |
| Jul 16, 2026 | Kimi K3 | 58.5 | PerceptionBench |
3 system cards report it, from 3 labs
Besides launch posts, we read the system cards, model cards and technical reports labs publish with a release. They report far more, including safety and frontier-risk tests.
| Date | Document | Score | As the document writes it | Section |
|---|---|---|---|---|
| Nov 20, 2026 | | 57.5 | PerceptionBench | Capabilities |
| Nov 9, 2026 | | 80.6 | PerceptionBench with tools, Muse Ember max reasoning | Capabilities |
| Jul 27, 2026 | Kimi K3 technical report | 58.5 | PerceptionBench (Table 2) | Capabilities |
One lab claim next to the leaderboard
Each score a lab reported for its own model, beside the same model's best score on the leaderboard. Gaps mostly come from different harnesses, effort levels and task subsets. 1 of them match, which usually means one side copied the other. Every benchmark.
More 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 11 checks pass. Every benchmark
Catalog record 2 pass
- Pass Task access The tasks are private.
- Pass Confidence Checked against a primary source.
Scores 4 pass 1 note
- Pass Independent runs Scores from the official leaderboard (16 entries), launch posts (6 scores) and system cards (3 scores).
- Pass Models scored 21 models we track.
- Pass Newest score The newest is from November 2026, for Inkling 2 Mini.
- Note Frontier models None of the top 10 has a score. The strongest tested is Claude Fable 5 (#12).
- Pass Headroom The best is 80.6%, under 85%, so it still separates models.
Leaderboard import 3 pass
- Pass Importer We read the board on github.com: 16 rows for models we track at the last read.
- Pass Last read November 18th, 2026.
- Pass Read worked The last read worked.
Consistency 2 pass
- Pass Lab claims The one pair agrees within 10 points.
- Pass Known issues No published audit found problems.