DeepSeekMath V2
DeepSeek-Math-V2 is a math-specialized LLM built on DeepSeek-V3.2-Exp-Base, trained to generate and verify step-by-step proofs. It uses a learned verifier as a reward model so the generator learns to fix its own reasoning, reaching gold-level scores on contests like IMO 2025, CMO 2024, and near-perfect Putnam 2024 with scaled test-time compute.
- Type
- Text
- Weights
- Open
More from DeepSeek
12
-
Coder
Text
AI model by DeepSeek
-
Deepseek 3.2
Text
DeepSeek V3.2 is the core, stable 3.2-series model designed as a strong general-purpose LLM. It builds on the…
-
5.2k DL
DeepSeek 3.2 Speciale
Text
DeepSeek-V3.2-Speciale is a 685B-parameter research-only variant of DeepSeek-V3.2 that pushes open-weight rea…
-
2.2M DL
DeepSeek OCR
Multimodal
LLM-centric OCR model using “Contexts Optical Compression” to explore visual-text compression and provide fas…
-
844k DL
DeepSeek OCR 2
Multimodal
Second-generation DeepSeek OCR model, “Visual Causal Flow,” aimed at more human-like visual encoding, with dy…
-
DeepSeek R1 Distill
Text
DeepSeek R1 Distill is a family of six small reasoning models released alongside DeepSeek R1, created by dist…
-
DeepSeek v2.5
Text
AI model by DeepSeek
-
DeepSeek V3
Text
AI model by DeepSeek
About DeepSeek
1
DeepSeek researches and develops general-purpose large language models, AI assistants, APIs, and open-source AI infrastructure.