DeepSeek Flash v4 Cybersecurity Test: Finds 24 CVEs but Loses on Cost-Efficiency to Luna
teortaxesTex · x · 2026-08-01
In the latest cybersecurity vulnerability hunting benchmark, DeepSeek Flash v4 demonstrated strong capabilities. The test results show that the model successfully identified 24 out of 32 CVEs (Common Vulnerabilities and Exposures), achieving a 75% pass@3 recall rate.
However, from a cost-efficiency perspective, DeepSeek's test cost $150.22. In contrast, the Luna model costs only $75.36 after repricing. Although DeepSeek finished just 1 CVE behind GPT-5.6-Sol and Kimi, Luna still maintains the lead on the Pareto frontier (the optimal balance of capability and cost) due to its lower price and consistency.
More from Models
- OpenAI stealth-testing mysterious internal model 'Aeon' in Codex desktop app and Dots — cedric_chee · 2026-10-16
- Argon edges out astra as daily driver, but Claude still wins heavy coding work — iruletheworldmo · 2026-10-16
- NVIDIA ships FP4-quantized Qwen3.8-Flash for local use, trending on Hugging Face — nvidia · 2026-10-16
- RoboVista benchmark: Gemini-4 Argon hits 81.9% on robotics understanding — berkeley_ai · 2026-10-16
- Dev reports decision model jev beats gpt-6-luna on quality in real training pipeline — genmon · 2026-10-16
- Google won't publish Argon 4 limits in AI Pro—users compare experiences — Consistent-Text-2662 · 2026-10-16