Build your own index
Every intelligence index is a weighted average of benchmark scores, and somebody picked the weights. Here, you pick them. See who comes out on top, and whether the ranking holds when the weights shift.
Start from
The benchmarks the most labs reported in 2026 (33 labs down to 20), minus any where top models already score past 92%. We use the version most recent models were scored on.
Tests and weights
Pick from 1,201 benchmarks with at least 8 scored models.
- Terminal-Bench 2.164 models
- SWE-Bench Pro97 models
- Humanity's Last Exam181 models
- LiveCodeBench v6110 models
- DeepSWE69 models
- MMLU-Pro295 models
- MMMU-Pro131 models
- AutomationBench72 models
Scores from
We use independent runs where they exist and fill the gaps with labs' own numbers.
Who comes out on top?
Kimi K3.1 Pro leads this mix with 82.9 and stays first under 80% of nearby weightings. Claude Fable 5.2 is second at 81.9. We rank the 85 models with scores on at least 4 of the 8 tests, from independent runs and labs’ own numbers.
- 82.9Kimi K3.1 Pro
- 81.9Claude Fable 5.2
- 79.3Claude Opus 5.5
- 77.5GLM-5.3
- 77.5
LongCat-2.5
- 76.0GPT-6 Astra
- 75.1Claude Opus 5
- 75.1GLM-5.3-Flash
- 74.9Claude Sonnet 5.5
- 74.7
Inkling 2 Small
- 74.0
Step 5 Preview
- 73.7
DeepSeek V4 Pro 0813
- 73.1Claude Fable 5
- 73.1
Gemini 3.8 Flash
- 72.9GPT-5.6 Sol
- 72.0
DeepSeek V4.1 Flash
- 71.9Kimi K3
- 71.5
Tencent Hy4 preview
- 71.2Claude Fable 5.1
- 71.1
Ling-3.1-flash
- 70.2
Qwen 3.8 Max
- 68.7Claude Opus 4.8
- 68.4GPT-5.6 Terra
- 68.3Beam 1.0
- 67.9
Muse Spark 1.1
- 67.0GPT-5.5
- 65.8GPT-5.6 Luna
- 65.8
DeepSeek-V4-Pro
- 65.7Claude Sonnet 5
- 65.4
Qwen3.8-Omni-Flash
- 65.1
Qwen3.8 Flash
- 64.5GLM-5.2
- 63.2
Gemini 3.6 Flash
- 61.9
Seed2.1 Pro
- 61.1
DeepSeek-V4-Flash
- 61.0Claude Opus 4.7
- 61.0
Qwen 3.8 27B
- 61.0
Tencent Hy3
- 60.4
Gemini 3.5 Flash
- 60.4
Intern-S2-397B
Sources: Official leaderboards, Epoch AI (CC BY 4.0) and labs’ model cards
| Rank | Model | Index | Tests | Best | Rank range | Worst | First |
|---|---|---|---|---|---|---|---|
| 1 | Kimi K3.1 Pro | 82.9 | 4 of 8 | 1 | 2 | 80% | |
| 2 | Claude Fable 5.2 | 81.9 | 4 of 8 | 1 | 2 | 21% | |
| 3 | Claude Opus 5.5 | 79.3 | 4 of 8 | 3 | 3 | ||
| 4 | GLM-5.3 | 77.5 | 4 of 8 | 4 | 5 | ||
| 5 | 77.5 | 4 of 8 | 4 | 6 | |||
| 6 | GPT-6 Astra | 76.0 | 4 of 8 | 5 | 7 | ||
| 7 | Claude Opus 5 | 75.1 | 4 of 8 | 5 | 12 | ||
| 8 | GLM-5.3-Flash | 75.1 | 4 of 8 | 6 | 15 | ||
| 9 | Claude Sonnet 5.5 | 74.9 | 4 of 8 | 7 | 15 | ||
| 10 | 74.7 | 4 of 8 | 7 | 16 | |||
| 11 | 74.0 | 5 of 8 | 8 | 17 | |||
| 12 | 73.7 | 4 of 8 | 8 | 16 | |||
| 13 | Claude Fable 5 | 73.1 | 5 of 8 | 8 | 19 | ||
| 14 | 73.1 | 4 of 8 | 9 | 18 | |||
| 15 | GPT-5.6 Sol | 72.9 | 5 of 8 | 8 | 18 | ||
| 16 | 72.0 | 6 of 8 | 6 | 31 | |||
| 17 | Kimi K3 | 71.9 | 5 of 8 | 14 | 20 | ||
| 18 | 71.5 | 5 of 8 | 15 | 20 | |||
| 19 | Claude Fable 5.1 | 71.2 | 4 of 8 | 15 | 22 | ||
| 20 | 71.1 | 5 of 8 | 12 | 23 | |||
| 21 | 70.2 | 6 of 8 | 18 | 22 | |||
| 22 | Claude Opus 4.8 | 68.7 | 5 of 8 | 20 | 25 | ||
| 23 | GPT-5.6 Terra | 68.4 | 5 of 8 | 20 | 26 | ||
| 24 | Beam 1.0 | 68.3 | 5 of 8 | 21 | 29 | ||
| 25 | 67.9 | 4 of 8 | 21 | 27 | |||
| 26 | GPT-5.5 | 67.0 | 5 of 8 | 23 | 28 | ||
| 27 | GPT-5.6 Luna | 65.8 | 5 of 8 | 25 | 32 | ||
| 28 | 65.8 | 5 of 8 | 26 | 33 | |||
| 29 | Claude Sonnet 5 | 65.7 | 5 of 8 | 25 | 31 | ||
| 30 | 65.4 | 4 of 8 | 25 | 32 | |||
| 31 | 65.1 | 4 of 8 | 28 | 33 | |||
| 32 | GLM-5.2 | 64.5 | 5 of 8 | 27 | 34 | ||
| 33 | 63.2 | 4 of 8 | 32 | 36 | |||
| 34 | 61.9 | 5 of 8 | 31 | 40 | |||
| 35 | 61.1 | 6 of 8 | 33 | 49 | |||
| 36 | Claude Opus 4.7 | 61.0 | 4 of 8 | 34 | 44 | ||
| 37 | 61.0 | 5 of 8 | 34 | 48 | |||
| 38 | 61.0 | 4 of 8 | 34 | 47 | |||
| 39 | 60.4 | 6 of 8 | 36 | 50 | |||
| 40 | 60.4 | 4 of 8 | 34 | 51 | |||
| 41 | GPT-5.4 | 60.3 | 6 of 8 | 35 | 48 | ||
| 42 | Kimi K2.6 | 59.8 | 5 of 8 | 34 | 53 | ||
| 43 | 59.6 | 5 of 8 | 34 | 51 | |||
| 44 | 59.5 | 4 of 8 | 36 | 48 | |||
| 45 | 59.2 | 4 of 8 | 36 | 50 | |||
| 46 | GLM-5.1 | 58.7 | 4 of 8 | 36 | 53 | ||
| 47 | 58.6 | 4 of 8 | 40 | 51 | |||
| 48 | 58.4 | 6 of 8 | 41 | 53 | |||
| 49 | 58.2 | 5 of 8 | 40 | 53 | |||
| 50 | 57.7 | 6 of 8 | 39 | 59 | |||
| 51 | 56.7 | 4 of 8 | 44 | 58 | |||
| 52 | 56.7 | 4 of 8 | 46 | 58 | |||
| 53 | Claude Opus 4.6 | 56.3 | 4 of 8 | 50 | 58 | ||
| 54 | 55.4 | 4 of 8 | 49 | 63 | |||
| 55 | 55.1 | 4 of 8 | 49 | 64 | |||
| 56 | 54.7 | 5 of 8 | 51 | 66 | |||
| 57 | 54.6 | 4 of 8 | 51 | 66 | |||
| 58 | Claude Sonnet 4.6 | 54.4 | 5 of 8 | 52 | 65 | ||
| 59 | Kimi K2.5 | 54.1 | 5 of 8 | 53 | 62 | ||
| 60 | 54.1 | 7 of 8 | 53 | 67 | |||
| 61 | 53.8 | 4 of 8 | 54 | 65 | |||
| 62 | 53.6 | 4 of 8 | 52 | 63 | |||
| 63 | 53.5 | 5 of 8 | 56 | 64 | |||
| 64 | GPT-5 | 51.7 | 4 of 8 | 60 | 68 | ||
| 65 | Claude Opus 4.5 | 51.5 | 4 of 8 | 62 | 68 | ||
| 66 | 51.1 | 5 of 8 | 61 | 69 | |||
| 67 | 50.9 | 4 of 8 | 61 | 70 | |||
| 68 | GPT-5.2 | 50.3 | 4 of 8 | 62 | 72 | ||
| 69 | o3 | 48.6 | 4 of 8 | 68 | 73 | ||
| 70 | 47.9 | 4 of 8 | 68 | 75 | |||
| 71 | 47.7 | 4 of 8 | 63 | 78 | |||
| 72 | 47.4 | 5 of 8 | 69 | 76 | |||
| 73 | Claude Sonnet 4.5 | 46.8 | 4 of 8 | 69 | 76 | ||
| 74 | 46.7 | 4 of 8 | 70 | 76 | |||
| 75 | 45.3 | 4 of 8 | 67 | 81 | |||
| 76 | LFM3-1.7B | 43.9 | 4 of 8 | 55 | 82 | ||
| 77 | 42.9 | 4 of 8 | 75 | 77 | |||
| 78 | 40.4 | 4 of 8 | 77 | 80 | |||
| 79 | Granite 4.2 30B | 39.0 | 4 of 8 | 78 | 82 | ||
| 80 | 38.7 | 4 of 8 | 77 | 83 | |||
| 81 | Claude Sonnet 4 | 38.5 | 4 of 8 | 77 | 82 | ||
| 82 | 37.6 | 4 of 8 | 78 | 82 | |||
| 83 | Kimi K2 | 32.7 | 4 of 8 | 83 | 85 | ||
| 84 | gpt-oss-120b | 32.5 | 4 of 8 | 82 | 85 | ||
| 85 | 30.1 | 4 of 8 | 83 | 85 |
How much do the weights matter?
In the default mix, 4 of the top 10 models can move three places or more when the weights shift.
How it works
We put every test on a 0 to 100 scale. Percentages stay as they are. Elo and other units run from the lowest model to the highest, and tests where lower is better get flipped. A model's index is the weighted average.
Where a model skipped a test, we predict that score from its other results and the test's difficulty, so skipping a hard test doesn't help and taking one doesn't hurt. Hover a model to see which of its scores are predicted.
Why weights matter
Two indexes built from the same scores can disagree on who's first, because the weights decide which tests count most. Published indexes change their weights between versions, and every change reorders some models.
So we rerank every model under 400 weightings near yours, each weight moved up or down by as much as half. Rank range shows each model's best and worst place, and First is the share of those weightings it wins.
What it leaves out
Independent runs cover far fewer new models than labs' own posts, so the board starts with both. Switch to Independent runs to leave lab numbers out, though the newest models may drop off.
Predictions assume a model's strengths carry across tests, so they miss a model that's strong at one thing and weak at another. Raise the minimum to lean on them less. And no average says how sure to be about any one score.