Arena

United States · AI

Arena evaluates AI models through comparisons and observations of how people use them in real tasks.

Product

The platform began with human preference comparisons, where users compare model responses, and expanded into factuality and agent-performance evaluation. Developers and researchers use its results to understand differences between models across tasks.

Agent Arena examines work such as coding, document analysis and creative tasks. Its Alignment Index adds measures of agent behavior, including actions outside a user’s permission and claims that unfinished work is complete, alongside measures of whether a model can accomplish the task.

Funding history

2 rounds · 2 with disclosed amounts
AnnouncedRoundAmountInvestors
Series B$200MKhosla Ventures — LeadLightspeed — Lead01 Advisors — ParticipantAcrew Capital — ParticipantAMP PBC — ParticipantAndreessen Horowitz (a16z) — ParticipantDell Technologies Capital — ParticipantEndeavor Catalyst — ParticipantFelicis — ParticipantPanos Madamopoulos — ParticipantQuantumLight — ParticipantSalesforce Ventures — ParticipantThe House Fund — Participant
Series A$150MAndreessen Horowitz (a16z) — Role not disclosedFelicis — Role not disclosedKleiner Perkins — Role not disclosedLaude Ventures — Role not disclosedLDVP — Role not disclosedLightspeed — Role not disclosedThe House Fund — Role not disclosedUC Investments — Role not disclosed
Sourcesarena.ai ↗