reasoning benchmark
MLS-Bench Lite
MLS-Bench Lite is the official 30-task subset of MLS-Bench for evaluating whether AI systems can invent generalizable and scalable machine learning methods across LLM pretraining and post-training, robotics, world models, computer vision, reinforcement learning, optimization, ML systems, and AI for Science.
Updated Aug 11, 2026
Models3
Model coverage3
MetricScore
EvidenceB
MLS-Bench Lite Ranking
Higher score ranks better on this benchmark.
| 01 | MA | 48.3% | 100.0% | 3 | C | |
| 02 | AC | 41.0% | 50.0% | 3 | C | |
| 03 | MA | 35.1% | 0.0% | 3 | C |
MLS-Bench Lite Score Distribution
A closer view of the leading scores on this benchmark.
MLS-Bench Lite Highlights
The leading models and scores on this benchmark.
What is MLS-Bench Lite?
What MLS-Bench Lite measures and how its scores work.
MLS-Bench Lite is the official 30-task subset of MLS-Bench for evaluating whether AI systems can invent generalizable and scalable machine learning methods across LLM pretraining and post-training, robotics, world models, computer vision, reinforcement learning, optimization, ML systems, and AI for Science.
Scores are shown in ratio. This benchmark is not independently verified and has an evidence level of B.
- Family
- MLS-Bench Lite
- Modality
- text
- Primary category
- reasoning
- Score direction
- higher
- LLMBoard eligible
- Yes
- Evaluation key
- mls-bench-lite|llm-stats-current
Benchmark scores retain their original unit. Overall score eligibility is shown separately.
FAQ
Common questions about MLS-Bench Lite.
Which model scores highest on MLS-Bench Lite?
Kimi K3 is currently ranked first with 48.3%.
What does MLS-Bench Lite measure?
MLS-Bench Lite is the official 30-task subset of MLS-Bench for evaluating whether AI systems can invent generalizable and scalable machine learning methods across LLM pretraining and post-training, robotics, world models, computer vision, reinforcement learning, optimization, ML systems, and AI for Science.
Is a higher score better?
Yes. Higher values rank better for this benchmark.
How many models are compared?
3 model results are currently shown.
Does this benchmark affect the overall score?
Yes. This benchmark can contribute to the current LLMBoard capability score.