OpenMathInstruct
by NVIDIA · open-source · Last verified 2026-03-17
OpenMathInstruct is a large-scale synthetic mathematics instruction dataset produced by NVIDIA, containing over 1.8 million math problem-solution pairs covering arithmetic, algebra, geometry, calculus, and competition mathematics. Solutions are generated by Mixtral models and verified for correctness to provide reliable step-by-step reasoning chains for training math-capable language models.
https://huggingface.co/datasets/nvidia/OpenMathInstruct-2 ↗B
B—Above Average
Adoption: B+Quality: AFreshness: ACitations: B+Engagement: F
Specifications
- License
- CC BY 4.0
- Pricing
- open-source
- Capabilities
- math-reasoning, instruction-tuning, chain-of-thought
- Integrations
- huggingface-datasets
- Use Cases
- math-model-training, reasoning-finetuning, competition-math
- API Available
- Yes
- Tags
- synthetic, math, instruction-tuning, reasoning, step-by-step
- Added
- 2026-03-17
- Completeness
- 100%
Index Score
64.2Adoption
73
Quality
85
Freshness
88
Citations
72
Engagement
0
Put AI to work for your business
Deploy this dataset alongside autonomous AaaS agents that handle tasks end-to-end — no babysitting required.