Tools Model Security Capability deepseek-v4-flash
Model record
DeepSeek V4 Flash: how it does on security work, and what it will answer
- Lab
- DeepSeek
- Released
- 2026-07-31
- Access
- open-weights
- Weights
- open
- Safeguards
- DeepSeek licence lists twelve use restrictions; cyber activity is absent from the list
- Score rows
- 1
- Confidence
- CONFIRMED
What happened
The cheapest measured pass on RealVuln: all 140 repositories at $0.0005 per 100 lines, F3 41.8 at 56.31% precision, $3.83 in total over three trials. A 120,000-line project costs about $0.62.
RealVuln 3.1.0, F3 (micro): 41.8 (independent; agentic harness (deepseek-v4-flash-agentic-v1), prompt sha256:45a1200d61e6; 140 of 140 repositories pinned by commit SHA (full corpus, TypeScript and JavaScript included); safeguards: open weights, no classifier; run 2026-09-11; CONFIRMED). strict F3 41.8; precision 56.31%, recall 40.67%; $3.83 for the run; $0.0005 per 100 lines; 3 runs; 231.8s wall clock a run. 188 successful runs over 140 repositories, so cost per run carries about 1.34 passes a repository. Refusal gate (CONFIRMED, observed 2026-09-11): Licence use restrictions name nothing on intrusion or exploitation.
Sources
- RealVuln dashboardraw.githubusercontent.com/kolega-ai/Real-Vuln-Benchmark/main…
- Refusal policy or licenceraw.githubusercontent.com/deepseek-ai/DeepSeek-V3/main/LICEN…
On YFarmX
- Reference pageyfarmx.com/ai/llms/deepseek-v4/
One record from the Model Security Capability, maintained by the Security Desk. Data: CSV · JSON ·RSS · CC BY 4.0 with attribution to YFarmX.Tracker updated · 19 September 2026