Open-weight models now match frontier cyber performance from just four months ago at a fraction of the cost

What happened

The British AI Security Institute reports that open-weight AI models, such as GLM-5.2 and DeepSeek V4-Pro, now perform at the same level as closed-source frontier models from only four months ago. At the start of 2025, open-weight models lagged by six to ten months in cyber capabilities but have rapidly closed that gap. These open models come at a significantly lower cost compared to proprietary frontiers running more advanced cyber tasks.

Why it matters

This narrowing performance difference puts powerful cyber capabilities in the hands of anyone using open-weight models. Organizations relying on closed frontier models no longer have a uniquely advanced edge that justifies the higher cost. The cost-effective open options accelerate access and adoption but come with a caveat: the British AI Security Institute found that existing safety layers on open models are mostly ineffective. That means defenders get less warning and fewer protections against emerging cyber threats. Open-weight models lower the bar to entry for sophisticated cyber tools, raising the risk environment without corresponding safety improvements.

What to watch next

Monitor how security teams adjust to the compressed timeframe for responding to cyber threats driven by open-weight models. Expect shifts in vendor positioning as cheaper open architectures capture more market share. Watch for development of new safety mechanisms that can keep pace with rapidly improving open models or regulators stepping in to manage this risk. The dynamic landscape will force operators to rethink reliance on closed-source advantages and prepare for faster-moving adversaries empowered by affordable AI.

AI Quick Briefs Editorial Desk