ARC-AGI
Integrity rank #52 of 65 · 85 models scored · top score 98.5 · GPT-6.1 Sol
unknown
Strongest on discrimination, weakest on contamination resistance. Caveat: the public test set does not rule out training-data contamination.
Reported scores — protocols may differ. How we compare
| # | Model | Score | Evidence | Measured |
|---|---|---|---|---|
| 1 | GPT-6.1 Sol | 98.5 | unverifiedT1 | 2026-09-29 |
| 2 | Gemini 3.8 Flash | 98.5 | unverifiedT1 | 2026-09-02 |
| 3 | Claude Fable 5 | 98.5 | unverified· optimizedT1 | 2026-06-09 |
| 4 | Claude Opus 5.5 | 98.5 | unverifiedT1 | 2026-09-22 |
| 5 | GPT-6 Astra | 98.5 | unverifiedT1 | 2026-09-03 |
| 6 | Gemini 3.1 Pro | 98 | unverified· optimizedT1 | 2026-02-19 |
| 7 | Claude Fable 5.1 | 97.5 | unverifiedT1 | 2026-09-01 |
| 8 | Claude Opus 5 | 97.5 | unverified· optimizedT1 | 2026-07-24 |
| 9 | GPT-5.6 Sol | 97.5 | unverified· optimizedT1 | 2026-07-09 |
| 10 | GPT-5.5 Pro | 96.5 | unverified· optimizedT1 | 2026-04-23 |
| 11 | GPT-5.6 Terra | 96.5 | unverified· optimizedT1 | 2026-07-09 |
| 12 | GPT-6 Sol | 95.5 | unverifiedT1 | 2026-09-22 |
| 13 | Gemini 3.7 Flash | 95.5 | unverified· optimizedT1 | 2026-08-13 |
| 14 | GPT-5.5 | 95 | unverified· optimizedT1 | 2026-04-23 |
| 15 | Kimi K3 | 94.5 | unverified· optimizedT1 | 2026-07-16 |
| 16 | DeepSeek V4.1 Flash | 94.5 | unverifiedT1 | 2026-09-09 |
| 17 | GPT-5.4 Pro | 94.5 | unverified· optimizedT1 | 2026-03-05 |
| 18 | Claude Opus 4.6 | 94 | unverified· optimizedT1 | 2026-02-05 |
| 19 | GPT-5.4 | 93.67 | unverified· optimizedT1 | 2026-03-05 |
| 20 | Claude Opus 4.7 | 93.5 | unverified· optimizedT1 | 2026-04-16 |
| 21 | Gemini 3.5 Flash | 92.5 | unverified· optimizedT1 | 2026-05-19 |
| 22 | Claude Opus 4.8 | 92.5 | unverified· optimizedT1 | 2026-05-28 |
| 23 | Gemini 3.6 Flash | 91.17 | unverified· optimizedT1 | 2026-07-21 |
| 24 | GLM-5.3-Flash | 91 | unverifiedT1 | 2026-08-20 |
| 25 | DeepSeek V4 Pro 0813 | 90.5 | unverified· optimizedT1 | 2026-08-13 |
| 26 | GPT-5.2 Pro | 90.5 | unverified· optimizedT1 | 2025-12-11 |
| 27 | Grok 4.7 | 90.17 | unverifiedT1 | 2026-09-21 |
| 28 | Grok 4.20 | 89.5 | unverified· optimizedT1 | 2026-02-17 |
| 29 | DeepSeek V4 Flash 0731 | 89 | unverified· optimizedT1 | 2026-07-31 |
| 30 | GPT-5.6 Luna | 88 | unverified· optimizedT1 | 2026-07-09 |
| 31 | Grok 4.6 | 87.5 | unverified· optimizedT1 | 2026-08-12 |
| 32 | Qwen 3.8 27B | 87.5 | unverifiedT1 | 2026-08-14 |
| 33 | Grok 4.5 | 87.17 | unverified· optimizedT1 | 2026-07-08 |
| 34 | GPT-6 Luna | 86.67 | unverifiedT1 | 2026-09-22 |
| 35 | Claude Sonnet 4.6 | 86.5 | unverified· optimizedT2 | 2026-02-17 |
| 36 | GPT-5.2 | 86.2 | unverified· optimizedT1 | 2025-12-11 |
| 37 | Gemini 3 Flash | 84.67 | unverified· optimizedT1 | 2025-12-17 |
| 38 | Inkling-Small | 84 | unverified· optimizedT1 | 2026-07-15 |
| 39 | Claude Opus 4.5 | 80 | unverified· optimizedT1 | 2025-11-24 |
| 40 | Inkling | 79.5 | unverified· optimizedT1 | 2026-07-15 |
| 41 | GLM-5.2 | 77 | unverified· optimizedT1 | 2026-06-16 |
| 42 | Gemini 3 Pro | 75 | unverified· optimizedT1 | 2025-11-18 |
| 43 | GPT-5.1 | 72.83 | unverified· optimizedT1 | 2025-11-13 |
| 44 | GPT-5 Pro | 70.2 | unverified· optimizedT1 | 2025-10-07 |
| 45 | Grok 4 | 66.67 | unverified· optimizedT2 | 2025-07-09 |
| 46 | GPT-5 | 65.67 | unverified· optimizedT1 | 2025-08-07 |
| 47 | Kimi K2.5 | 65.33 | unverified· optimizedT2 | 2026-01-27 |
| 48 | MiniMax-M2.5 | 63.67 | unverified· optimizedT2 | 2026-02-12 |
| 49 | Claude Sonnet 4.5 | 63.67 | unverified· optimizedT2 | 2025-09-29 |
| 50 | GPT-5.4 Mini | 63.67 | unverified· optimizedT1 | 2026-03-17 |
| 51 | o3 | 60.83 | unverified· optimizedT1 | 2025-04-16 |
| 52 | o3-pro | 59.33 | unverified· optimizedT1 | 2025-06-10 |
| 53 | o4-mini | 58.67 | unverified· optimizedT1 | 2025-04-16 |
| 54 | DeepSeek-V3.2 | 57 | unverified· optimizedT2 | 2025-12-01 |
| 55 | GPT-5 mini | 54.33 | unverified· optimizedT2 | 2025-08-07 |
| 56 | Gemini 3.5 Flash-Lite | 53.5 | unverified· optimizedT1 | 2026-07-21 |
| 57 | GPT-5.4 Nano | 51.5 | unverified· optimizedT1 | 2026-03-17 |
| 58 | Grok 4 Fast | 48.5 | unverified· optimizedT2 | 2025-09-19 |
| 59 | Claude Haiku 4.5 | 47.67 | unverified· optimizedT2 | 2025-10-15 |
| 60 | GLM-5 | 44.67 | unverified· optimizedT2 | 2026-02-11 |
| 61 | Gemini 2.5 Pro (Jun 2025) | 41 | unverified· optimizedT1 | 2025-06-05 |
| 62 | Claude Sonnet 4 | 40 | unverified· optimizedT1 | 2025-05-22 |
| 63 | Claude Opus 4 | 35.67 | unverified· optimizedT1 | 2025-05-22 |
| 64 | o3-mini | 34.5 | unverified· optimizedT1 | 2025-01-31 |
| 65 | Gemini 2.5 Flash (May 2025) | 33.33 | unverified· optimizedT1 | 2025-05-20 |
| 66 | Gemini 2.5 Pro (Mar 2025) | 33 | unverified· optimizedT1 | 2025-03-31 |
| 67 | Gemini 2.5 Flash (Apr 2025) | 32.3 | unverified· optimizedT1 | 2025-04-17 |
| 68 | o1 | 30.7 | unverified· optimizedT1 | 2024-12-17 |
| 69 | Claude 3.7 Sonnet | 28.6 | unverified· optimizedT1 | 2025-02-24 |
| 70 | DeepSeek-R1 (May 2025) | 21.21 | unverified· optimizedT1 | 2025-05-28 |
| 71 | GPT-5 nano | 20.71 | unverified· optimizedT2 | 2025-08-07 |
| 72 | o1-preview | 18 | unverified· optimizedT1 | 2024-09-12 |
| 73 | Grok-3 mini | 16.5 | unverified· optimizedT1 | 2025-06-24 |
| 74 | DeepSeek-R1 | 15.8 | unverified· optimizedT1 | 2025-01-20 |
| 75 | o1-mini | 14 | unverified· optimizedT1 | 2024-09-12 |
| 76 | Qwen3-235B-A22B-Instruct (Jul 2025) | 11 | unverified· optimizedT2 | 2025-07-25 |
| 77 | GPT-4.5 | 10.3 | unverified· optimizedT1 | 2025-02-27 |
| 78 | Grok 3 | 5.5 | unverified· optimizedT1 | 2025-04-09 |
| 79 | GPT-4.1 | 5.5 | unverified· optimizedT1 | 2025-04-14 |
| 80 | Magistral Small 1.0 | 5 | unverified· optimizedT2 | 2025-06-10 |
| 81 | GPT-4o (Nov 2024) | 4.5 | unverified· optimizedT1 | 2024-11-20 |
| 82 | Llama 4 Maverick | 4.38 | unverified· optimizedT1 | 2025-04-05 |
| 83 | GPT-4.1 mini | 3.5 | unverified· optimizedT1 | 2025-04-14 |
| 84 | Llama 4 Scout | 0.5 | unverified· optimizedT1 | 2025-04-05 |
| 85 | GPT-4.1 nano | 0 | unverified· optimizedT1 | 2025-04-14 |
Continue from this leaderboard into the tightest head-to-head reads.
Weighted composite. Each dimension opens to the source-backed sub-signals it is computed from.
Claims drawn from cited facts, not live model generation.
The recorded metadata does not specify what this benchmark measures.
1 cited fact
This benchmark has scores from 85 tracked models, with a recorded score range spanning 98.5.
2 cited facts
Its Benchmark Integrity Index score is 71, with grade B and rank 52 out of 65 tracked benchmarks.
4 cited facts
It is flagged as saturated. The top recorded score is 98.5, with 1.5 points of headroom. The computed top-score cluster contains 11 models.
4 cited facts
The harness classification is consistent harness. Test-set privacy is recorded as unknown. The recorded contamination-window ratio is 1.0.
3 cited facts
ARC-AGI: ARC-AGI as reported in Epoch AI's Capabilities Index CSV.
GPT-6.1 Sol leads ARC-AGI at 98.5. The full leaderboard above lists every recorded measurement, not just the headline number.
85 models have recorded scores on ARC-AGI, spanning a score spread of 98.5.
tensor.news grades ARC-AGI B for integrity (score 71/100), ranking #52 of 65 benchmarks we assess across discrimination, saturation, contamination resistance, harness comparability, and freshness.
Yes — top scores are clustering near the ceiling (saturation ratio 0.13), so ARC-AGI no longer separates leading models well. Treat small gaps at the top with caution.
Its test set is unknown, so contamination risk is on the table: a high score may partly reflect training-data overlap rather than capability.
Scores are reported under a consistent harness, so comparisons on ARC-AGI are reasonably apples-to-apples.
Every ARC-AGI measurement is source-backed and tagged reproduced, self-reported, or unverified; the underlying record cites epoch.ai/data/eci_benchmarks.csv.