Your team’s benchmarks
Build a private benchmark from the work your team already does. Compare models on quality, cost, and speed, choose a baseline, and when a new model comes out, run the same test to see whether switching is worth it.
Create my benchmark See a sample benchmark
Free to set up. First comparison US$299. No organisation needed to start; share a benchmark with your team when you are ready. How it works and pricing
How team benchmarks work
- Create a benchmark from five or more real cases: a prompt, its input, and what a good response must get right.
- Share it with your team. Members can read the reports and add cases. Only the owner can change the baseline, export, share or delete it.
- Re-test when models change. “Test a new model” reuses your baseline’s results and shows what improved or regressed.
Team prompt library
Keep the prompts your team already uses in one private place, so everyone starts from a good one, and pick any of them as a case when you build a benchmark.
- Create a team. Give it a name. It gets its own private page.
- Add prompts. From the review step when you share prompts, from My library, or by saving any public prompt to your team.
- Invite your team. Invite them by email or send a link. They sign in, join, then search and copy what they need.
Private by design
- Team prompts and team benchmarks are only visible to signed-in team members. Anyone else gets a “page not found”.
- They never appear in public search, the prompt library, the sitemap, or AI and search-engine files.
- The team owner can remove members and delete prompts or the whole team.
- Free plan: up to 5 members and 50 prompts per team, and one team you own per free account. Paid plans with more teams are coming soon.
