reasoning benchmark
RSI Index
The RSI (Recursive Self-Improvement) Index is OpenAI's aggregate metric across a bundle of internal AI-research evaluations, including debugging research systems, optimizing kernels and training recipes, and improving other models, measuring progress toward recursive self-improvement.
Updated Aug 7, 2026
Published models3
Registry coverage3
MetricScore
EvidenceB
RSI Index leaderboard
Sorted by the source-provided rank. Higher score is better according to the registry.
| 1 | OP | 57.9% | 100.0% | 3 | C | |
| 2 | OP | 56.3% | 50.0% | 3 | C | |
| 3 | OP | 41.9% | 0.0% | 3 | C |
Score distribution
Top published rows on the benchmark's original scale.
RSI Index highlights
The top published results on this benchmark's own scale.
What is RSI Index?
Definition and scoring fields from the benchmark registry.
The RSI (Recursive Self-Improvement) Index is OpenAI's aggregate metric across a bundle of internal AI-research evaluations, including debugging research systems, optimizing kernels and training recipes, and improving other models, measuring progress toward recursive self-improvement.
Scores are shown in ratio. The current registry marks this benchmark as not independently verified with evidence level B.
- Family
- RSI Index
- Modality
- text
- Primary category
- reasoning
- Score direction
- higher
- LLMBoard eligible
- No
- Evaluation key
- rsi-index|llm-stats-current
Source-native results are preserved. Eligibility for the overall LLMBoard score is a separate policy decision.
FAQ
Common questions about RSI Index.
Which model scores highest on RSI Index?
GPT-5.6 Sol is currently ranked first with 57.9%.
What does RSI Index measure?
The RSI (Recursive Self-Improvement) Index is OpenAI's aggregate metric across a bundle of internal AI-research evaluations, including debugging research systems, optimizing kernels and training recipes, and improving other models, measuring progress toward recursive self-improvement.
Is a higher score better?
Yes. Higher values rank better for this benchmark.
How many models are compared?
3 unique published model results are currently shown.
Does this benchmark affect the overall score?
This benchmark is preserved as source-native evidence but is not eligible for the current overall score.