Promptfoo was acquired by OpenAI in March 2026. If you need a vendor-neutral LLM evaluation and security platform, here are the best alternatives.
246 attack plugins, 145 scorers, 88 providers, compliance dashboard, LLM firewall. Full SaaS + self-hosted. Open source MIT.
Best for: Teams that need eval + security + compliance in one platform
Python-native eval with 50+ metrics and 20+ attack methods (DeepTeam). 12.8K stars, 400K+ downloads. Confident AI from $19.99/seat.
Best for: Python-only teams wanting pytest integration and growing red team features
Limitations: Python only, ~20 attacks (vs 232), no firewall/gateway/prompt IDE
Best-in-class LLM tracing and observability (YC W23). 100+ providers via LiteLLM. No red teaming or built-in eval.
Best for: Teams focused purely on LLM observability and tracing
Limitations: Zero attack plugins, no built-in eval scorers, no compliance
EU-focused AI red teaming with 40-50 probes and dynamic multi-turn agents. SOC 2 Type II certified.
Best for: EU enterprises needing adaptive red teaming with SOC 2 certification
Limitations: 40-50 probes, 10-15 scorers, no firewall/tracing/gateway
Closed-source eval platform with polished UX. No attack plugins or self-hosting.
Best for: Teams wanting simple eval-only workflows
Limitations: Closed source, no security testing, no self-host
NVIDIA's open-source LLM vulnerability scanner. CLI only, 37+ probes.
Best for: Security researchers wanting CLI-based probing
Limitations: CLI only, no eval, no dashboard, 37 probes
Databricks' open-source ML lifecycle platform with basic LLM eval (~12 scorers).
Best for: Teams already invested in Databricks ecosystem
Limitations: ~12 scorers, no security testing, SaaS requires Databricks
OpenAI's built-in evaluation. Free but locked to OpenAI models only.
Best for: Teams using only OpenAI models
Limitations: OpenAI only, no red teaming, vendor locked
Start free. No credit card required. Migrate in minutes.