Contamination MapText / LLM / Model
DeepSeek-V4 (Pro / Flash)
- Organisation
- DeepSeek-AI
- Family
- DeepSeek · version V4
- Released
- 2026-04-24 (DeepSeek API Docs news entry 'DeepSeek-V4 Preview Release 2026/04/24', https://api-docs.deepseek.com/news/news1226; arXiv v1 2026-04-26)
- Checkpoints
- deepseek-ai/DeepSeek-V4-Pro-Base, deepseek-ai/DeepSeek-V4-Pro, deepseek-ai/DeepSeek-V4-Flash-Base, deepseek-ai/DeepSeek-V4-Flash, deepseek-ai/DeepSeek-V4-Flash-0731, deepseek-ai/DeepSeek-V4-Pro-0813
- Weights
- Open · MIT
- Links
- PaperModel card
Training data disclosure: Vague
Level determined for 0% of benchmark cells.
- TestTest data seen0
- SeenSeen, split unclear0
- TrainTraining data seen0
- Related data seen0
- NoNot seen0
- ?Unknown15
“Combining all the above, our pre-training corpus comprises more than 32T tokens, containing mathematical contents, codes, web pages, long documents, and other high-quality categories.”
Other declared training data
-
web-sourced data (on top of the DeepSeek-V3 pre-training data) (broad web data: may contain benchmark test sets) · split: unknown
Evidence
“On top of the pre-training data of DeepSeek-V3, we endeavor to construct a more diverse and higher-quality training corpus with longer effective contexts.”
arxiv.org · paper · Section 4.1 · retrieved 2026-10-15 “For web-sourced data, we implement filtering strategies to remove batched auto-generated and templated content”
arxiv.org · paper · Section 4.1 · retrieved 2026-10-15 -
long documents (scientific papers, technical reports) (broad web data: may contain benchmark test sets) · split: unknown
Scientific papers can quote benchmark items; source collections not named.
Evidence
“we place a particular emphasis on long-document data curation, prioritizing scientific papers, technical reports, and other materials that reflect unique academic values.”
arxiv.org · paper · Section 4.1 · retrieved 2026-10-15 -
agentic data in mid-training (not yet matched to a dataset) · split: unknown
Evidence
“we further enhance the coding capabilities of DeepSeek-V4 series by incorporating agentic data during the mid-training phase.”
arxiv.org · paper · Section 4.1 · retrieved 2026-10-15
Benchmark cells
Classic LLM benchmarks
Unknown MMLU (Measuring Massive Multitask Language Understanding) · training data not disclosed in enough detail
Unknown GSM8K (Grade School Math 8K) · training data not disclosed in enough detail
Unknown MATH (Mathematics Aptitude Test of Heuristics) · training data not disclosed in enough detail
Unknown HumanEval · training data not disclosed in enough detail
Unknown ARC-Challenge (AI2 Reasoning Challenge, Challenge Set) · training data not disclosed in enough detail
Unknown HellaSwag · training data not disclosed in enough detail
Unknown TruthfulQA · training data not disclosed in enough detail
Unknown WinoGrande · training data not disclosed in enough detail
Recent LLM benchmarks
Unknown GPQA Diamond · training data not disclosed in enough detail
Unknown MMLU-Pro · training data not disclosed in enough detail
Unknown LiveCodeBench (code generation, release_v6) · training data not disclosed in enough detail
Unknown SWE-bench Verified · training data not disclosed in enough detail
Unknown AIME 2024 · training data not disclosed in enough detail
Unknown AIME 2025 · training data not disclosed in enough detail
Unknown MATH-500 · training data not disclosed in enough detail
Notes
V4-Flash pre-trained on 32T tokens, V4-Pro on 33T. DeepSeek-V4-Flash-0731 and DeepSeek-V4-Pro-0813 are the 'official release' versions superseding the April preview (post-training updates, no new data information). No cutoff or decontamination statement found. DeepSeek-V4-Flash-Vision-Exp (2026-08) not covered.
Sources
- DeepSeek-V4: Towards Highly Efficient Million-Token Context Intelligence · paper · retrieved 2026-10-15
- DeepSeek-V4-Pro model card · model card · retrieved 2026-10-15
- DeepSeek-V4-Flash model card · model card · retrieved 2026-10-15
- DeepSeek-V4-Pro-0813 model card · model card · retrieved 2026-10-15
- DeepSeek-V4-Flash-0731 model card · model card · retrieved 2026-10-15