Primary evidence
- DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language ModelsS1 · Primary research · Paper · 2024-02-05
- DeepSeekMath code, evaluation and model release repositoryS1 · Primary research · Repository
DeepSeekMath couples large-scale math pretraining with GRPO
The model continued pretraining on 120B math-related tokens and then used supervised and reinforcement-learning stages for mathematical reasoning.
What happened?
DeepSeekMath-7B reached 51.7% on MATH without external tools or voting and introduced Group Relative Policy Optimization (GRPO).
Why it matters
It linked open math-model scaling with a verifier-friendly RL recipe that influenced subsequent reasoning systems.
Significance rationale: DeepSeekMath paired a strong open mathematical corpus pipeline with GRPO, a reinforcement-learning method that later became influential in reasoning-model training.
Historical context
Earlier developments
Later developments
—
Related milestones
—
Evidence & status at a glance
- Significance
- H2 · Field milestone
- Verification
- Paper released
- Evidence level
- E3 · Inspectable artifact
- AI role
- Computation only
- Formal assurance
- Artifact available
Contribution & verification
- Mathematical novelty
- Benchmark result
- Interface
- Informal, Symbolic
- AI contribution
- The trained model performs competition-level mathematical reasoning.
- Human contribution
- Researchers built the data pipeline, model training recipe and GRPO optimization method.
- Verification
- Paper released · E3 · Inspectable artifact · Artifact available
- Last reviewed
- 2026-09-11
Technical details
- Systems
- deepseekmath
- Organizations
- deepseek
- Problems
- math-benchmark
- Methods
- grpo, continued-pretraining, reinforcement-learning
Sources & artifacts
- DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language ModelsS1 · Primary research · Paper · accessed 2026-09-10
- DeepSeekMath code, evaluation and model release repositoryS1 · Primary research · Repository · accessed 2026-09-11
- DeepSeekMath evaluation code and model linksArtifact · Code
Verification history
- 2024-02-05 · Paper released · E2 · Research record
The Chronicle records this event from the cited primary or official source; the status does not imply independent Chronicle verification.
- 2026-09-11 · Paper released · E3 · Inspectable artifact
This corpus refresh added a directly inspectable public artifact maintained by the authors or releasing organization. E3 records artifact-level inspectability and does not imply independent verification of the underlying scientific or mathematical claim.
Corrections
No corrections recorded.