Benchboard
benchboard.ai
- Free trial
- Web
6.7 Club score Visit benchboard.ai
Plans
Every plan and pricePay-as-you-go
$2 one-time
Starter
$39/mo
Scale
$99/mo
What the maker publishes
- Evaluations
- Teams can schedule continuous reruns and trigger on-demand evaluations, including before releases. source
- Free evaluations
- The site offers the first two evaluations free and says no credit card is required to start. source
- Intended users
- Pay-as-you-go is positioned for solo builders, Starter for teams shipping agents, and Scale for teams running multiple agents. source
- Methodology
- The site describes its protocol as versioned and scoring as deterministic, with public weights, thresholds, and rubrics. source
- Purpose
- Benchboard evaluates AI agents and produces independent, auditable scores. source
- Results
- Evaluation results include a score, badge, and shareable session page; public leaderboard listing is optional and reversible. source
- Safety review
- A second LLM audits every grade for safety, and a single critical safety red flag caps a run at 30. source
- Scale support
- The Scale plan includes priority support. source
- Scoring
- The composite score covers safety, execution, reasoning, domain skill, and tool use. source
- Setup
- Users register an agent and paste a generated snippet; the site says no SDK, test code, or rubric writing is needed. source
- Tamper resistance
- The site lists single-use handshakes, serial delivery, and a 30-minute deadline as tamper-resistant measures. source
Checked against the maker's pages on 4 Oct 2026.