Claude Opus 5 tops a prompt-injection robustness chart and drops attacks near zero
EricBuess · x · 2026-07-25
The screenshot shows a chart titled “Indirect prompt injection robustness: Gray Swan IPI benchmark” comparing several models.
The visible takeaway is that Opus 5 sits at the low end of the attack-success bars, with the chart suggesting it is the hardest model to successfully prompt inject among the models shown. The quoted text also says that when strong alignment, prompt-injection probes, and Claude Code Auto Mode are layered together, attack success drops to about 0.
The post frames this as a notable jump in real-world robustness rather than just another benchmark win.
Related event: Claude Opus 5 Sets New Record in Prompt Injection Resistance(2 posts)→
More from Models
- Rumor claims Kimi faked performance by serving Claude; DeepSeek new model surprises in evals — realsohamparekh · 2026-09-11
- GPT-5.6 writes well but is instantly forgettable, user complains — BasedRaddka · 2026-09-11
- Opus Refuses Protein Research Codebase Over 'Safety' Concerns, Dev Considers Rolling His Own — josephdviviano · 2026-09-11
- User Hails Unconfirmed 'DeepSeek 4.1 Flash' as an Inflection Point in LLMs — himanshustwts · 2026-09-11
- Terminal Bench v4: GLM-5.3 Leads at 41.9%, Kimi-K3 Underwhelms at 12.6% — Ok_Warning2146 · 2026-09-11
- GPT-6 Astra beats Factorio with enemies in 44 in-game hours at ~$4,500 API cost — liminal_bardo · 2026-09-11