DeepSeek-Math-V2
DeepSeek's math reasoning large model — IMO gold-medal level
DeepSeek's math reasoning large model at IMO gold-medal level.
Open foundation models — the base layers everything builds on.
Key features
- Self-verifiable mathematical reasoning model
- Trains an LLM verifier for theorem proving, used as the reward model
- Gold-level scores on IMO 2025; 118 of 120 on Putnam 2024
- Built on DeepSeek-V3.2-Exp-Base
Good for
- Mathematical theorem proving research
- Verifiable reasoning for hard math problems
Stack & tags
maththeorem-provingllmopen-weightsself-verificationreasoningrl