Support
Help & FAQ
Quick answers for getting started with BenchmarkFree. For legal or privacy matters, see our policy pages or email us directly.
Frequently asked questions
- Do I need an account to use BenchmarkFree?
- You can browse public prompts as a guest. To run evaluations, save models, or use private prompts, sign in with email and password.
- How do I run my first Quick Eval?
- Go to Quick Eval, select candidate models and a judge, pick a prompt, then click Start Eval. See How it works for a step-by-step guide.
- What is the judge model?
- The judge is a separate LLM that scores candidate outputs on configured dimensions (e.g. correctness, robustness). You choose it from the same catalog as candidate models; it must not be the same model as a candidate in the same run.
- Are my API keys safe?
- Saved API keys are encrypted at rest (AES-based Fernet) and transmitted over TLS. They are used only to call providers you select when you start a run, and are not visible to other users. See the Privacy Policy for full details.
- Why did scoring fail?
- Scoring can fail if the judge provider is unreachable, rate-limited, or misconfigured. Generation may still succeed — check the error on the result card and verify your judge model credentials and provider status.
- Will my results appear on the public leaderboard?
- Only eligible runs from public prompts that complete scoring may contribute aggregated scores. Your email and raw prompts are not published. Use private prompts to avoid public leaderboard inclusion.
- How do I delete my account?
- Email [email protected] from your registered address with subject "Account deletion request". We aim to complete deletion within 30 days. See Privacy Policy §10–11.
- Is BenchmarkFree free?
- The platform is free to use. You may incur charges from third-party model providers (OpenAI, Anthropic, etc.) when you use your own API keys or billable platform models.
Still need help?
Email us at [email protected]