Leaderboard bars
3 ranked models · higher is better
View accessible chart data
| Model | Rank | Provider | Score |
|---|---|---|---|
| GPT-5.6 Sol | #1 | OpenAI | 7.8% |
| GPT-5.6 Terra | #2 | OpenAI | 0.8% |
| GPT-5.6 Luna | #3 | OpenAI | 0.2% |
Loading benchmark header
Loading benchmark registry
Reasoning · Benchmark profile
An interactive reasoning benchmark testing exploration, world modelling, goal discovery, planning and adaptive execution in novel environments.
Data verified 15 July 2026 · Methodology 1.2.0
Benchmark score on ARC-AGI-3
GPT-5.6 Sol leads at 7.8%, followed by GPT-5.6 Terra (0.8%) and GPT-5.6 Luna (0.2%).
Visual analysis
Three views of the same sourced leaderboard: model placement, the score curve, and descriptive provider averages. These visuals never alter ranking calculations.
3 ranked models · higher is better
| Model | Rank | Provider | Score |
|---|---|---|---|
| GPT-5.6 Sol | #1 | OpenAI | 7.8% |
| GPT-5.6 Terra | #2 | OpenAI | 0.8% |
| GPT-5.6 Luna | #3 | OpenAI | 0.2% |
Shows how quickly scores separate down the field
| Model | Rank | Provider | Score |
|---|---|---|---|
| GPT-5.6 Sol | #1 | OpenAI | 7.8% |
| GPT-5.6 Terra | #2 | OpenAI | 0.8% |
| GPT-5.6 Luna | #3 | OpenAI | 0.2% |
Average within the visible cohort · descriptive only
| Provider | Models | Average score |
|---|---|---|
| OpenAI | 3 | 2.9% |
One best score per model · higher is better
| Rank | Model | Provider | License | Score |
|---|---|---|---|---|
| #1 | GPT-5.6 Sol gpt-5.6-sol | OpenAI | closed | 7.8% |
| #2 | GPT-5.6 Terra gpt-5.6-terra | OpenAI | closed | 0.8% |
| #3 | GPT-5.6 Luna gpt-5.6-luna | OpenAI | closed | 0.2% |
The top of this snapshot is led by GPT-5.6 Sol at 7.8%; third place is 7.6 points behind. The top-3 spread is 7.6 points.
About ARC-AGI-3
An interactive reasoning benchmark testing exploration, world modelling, goal discovery, planning and adaptive execution in novel environments. 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
An interactive reasoning benchmark testing exploration, world modelling, goal discovery, planning and adaptive execution in novel environments.
GPT-5.6 Sol by OpenAI currently leads with 7.8%.
3 models in the LuminaBench cohort have a qualifying score on this benchmark.
No. This benchmark is display-only and does not enter the overall Lumina composite.
Related
Rolling · 96 results · Weighted 2.5%
Humanity's Last ExamRolling · 105 results · Weighted 2.5%
ARC-AGI-22 · 10 results · Weighted 2.5%
LiveBenchRolling · 17 results · Weighted 2.5%
BenchLM Reasoning priorbench-align-v5.1 · 18 results · Weighted 1.1%
SuperGPQARolling · 8 results · Weighted 2.5%