Reasoning · Benchmark profile
A second-generation abstract reasoning benchmark based on novel grid transformation tasks.
Data verified 15 July 2026 · Methodology 1.2.0
Benchmark score on ARC-AGI-2
GPT-5.5 leads at 85%, followed by Gemini 3.1 Pro Preview (77.1%) and Gemini 3.5 Flash (72.1%).
Visual analysis
Three views of the same sourced leaderboard: model placement, the score curve, and descriptive provider averages. These visuals never alter ranking calculations.
7 ranked models · higher is better
| Model | Rank | Provider | Score |
|---|---|---|---|
| GPT-5.5 | #1 | OpenAI | 85% |
| Gemini 3.1 Pro Preview | #2 | 77.1% | |
| Gemini 3.5 Flash | #3 | 72.1% | |
| GPT-5.2 Pro | #4 | OpenAI | 54.2% |
| Gemini 3 Pro Deep Think | #5 | 45.1% | |
| Muse Spark | #6 | Meta | 42.5% |
| Gemini 3 Pro | #7 | 31.1% |
Shows how quickly scores separate down the field
| Model | Rank | Provider | Score |
|---|---|---|---|
| GPT-5.5 | #1 | OpenAI | 85% |
| Gemini 3.1 Pro Preview | #2 | 77.1% | |
| Gemini 3.5 Flash | #3 | 72.1% | |
| GPT-5.2 Pro | #4 | OpenAI | 54.2% |
| Gemini 3 Pro Deep Think | #5 | 45.1% | |
| Muse Spark | #6 | Meta | 42.5% |
| Gemini 3 Pro | #7 | 31.1% |
Average within the visible cohort · descriptive only
| Provider | Models | Average score |
|---|---|---|
| OpenAI | 2 | 69.6% |
| 4 | 56.3% | |
| Meta | 1 | 42.5% |
One best score per model · higher is better
| Rank | Model | Provider | License | Score |
|---|---|---|---|---|
| #1 | GPT-5.5 gpt-5.5 | OpenAI | closed | 85% |
| #2 | Gemini 3.1 Pro Preview gemini-3.1-pro-preview | closed | 77.1% | |
| #3 | Gemini 3.5 Flash gemini-3.5-flash | closed | 72.1% | |
| #4 | GPT-5.2 Pro gpt.5.2.pro | OpenAI | closed | 54.2% |
| #5 | Gemini 3 Pro Deep Think gemini-3-pro-deep-think | closed | 45.1% | |
| #6 | Muse Spark muse-spark | Meta | closed | 42.5% |
| #7 | Gemini 3 Pro gemini-3-pro | closed | 31.1% |
The top of this snapshot is led by GPT-5.5 at 85%; third place is 12.9 points behind. The top-7 spread is 53.9 points.
About ARC-AGI-2
A second-generation abstract reasoning benchmark based on novel grid transformation tasks. Results stay tied to the exact model variant and evaluation system. Multiple systems for the same model use the best published score on this page; overall Lumina scoring uses the median of ranking-eligible rows.
Open benchmark source ↗FAQ
A second-generation abstract reasoning benchmark based on novel grid transformation tasks.
GPT-5.5 by OpenAI currently leads with 85%.
7 models in the LuminaBench cohort have a qualifying score on this benchmark.
Yes. This family is ranking-weighted at 2.5% of its capability category. GPT-5.5 is overall #8.
Related
