Rankings/Arena
Arena logo

Arena

NativeAI NativePrivateUpdated November 15, 2026

The AI benchmarking community that shapes how frontier models are evaluated and improved.

Try it free →
Website
arena.ai
Founded
2022
CEO
Anastasios Angelopoulos
HQ
San Francisco, CA, United States
Stage
Private
Revenue
Not disclosed
Total Funding
$450M
Employees
201-500
Flag a data issue
AI InfrastructureData & Analytics
57.4Bluescore
Momentum67.2
Market MapDeep Tech
Rank change (7d)▲ 44

Executive Summary

Updated Nov 15, 2026

What they do

  • Runs Arena, a public platform where users vote between anonymous paired AI models
  • Publishes crowd-sourced leaderboards for text, code, web dev, search, image, video and document models
  • Sells commercial evaluation services, mainly to labs building frontier models

Why it matters

  • Its leaderboard is among the most watched signals of frontier model quality
  • Reports about 60 million monthly conversations, a large base of real-use preference data
  • Grew fast to a reported $100M annualized revenue by June 2026

Best fit

  • Model labs needing pre-release human-preference testing on real user prompts
  • Enterprise AI teams comparing frontier models before choosing vendors
  • Product teams tracking how coding, search and multimodal models rank

Watch-outs

  • Preference votes can reward style over accuracy, a key point in the Leaderboard Illusion critique
  • Lab customers paying for evaluations raise neutrality and conflict-of-interest questions
  • Public rankings may not reflect your own domain, data or compliance needs

Products & Offerings

Arena (public leaderboard platform)

Free web platform where users compare anonymous model responses side by side, and their votes feed public leaderboards for text, vision, code, search and more.

PlatformDeveloperFlagship2023

AI Evaluations (commercial)

Paid evaluation service that gives model developers structured human-preference testing and analysis from Arena's community before and after release.

EnterpriseEnterpriseHigh Growth2025

Specialized Arenas (WebDev, Search, Document, Video Edit)

Separate leaderboards that rate models on specific tasks such as web development, search, document handling and video editing; Document and Video Edit arenas launched in March 2026.

PlatformDeveloperLive

Arena Expert

Leaderboard built from harder, expert-level prompts to separate frontier models on demanding tasks.

DataResearchLive

Alignment Index

Benchmark launched in October 2026 that scores how AI agents behave on real tasks, not just how capable they appear.

DataEnterpriseEmerging2026

Latest News

Analyst VerdictEvaluate

Updated Nov 15, 2026

Arena is the best-known human-preference benchmark for frontier AI and is well funded, with about $450M raised, a 2026 Series B co-led by Lightspeed and Khosla, and fast revenue growth. Buyers should treat its public leaderboards as one input among several, not as a final answer. Commercial evaluations should be tested against your own tasks, with documented methods and clear independence safeguards.

Signal Provenance

Latest observation date and source for each signal category

Signal CategoryLast UpdatedSource
Funding Signals
Last funding round, Funding stage
Nov 11, 2026Cited web research
Hiring Velocity
Headcount, Headcount growth
Nov 15, 2026Research agents (estimate)
Technical Output
Research papers, GitHub activity
Nov 13, 2026Hugging Face Hub + news feed + research agents
News Sentiment
News mentions, News sentiment
Nov 15, 2026Verified news feed
Partnership Signals
Enterprise partnerships, Distribution partners
Oct 31, 2026Research agents (estimate)
Revenue Indicators
ARR estimate, Revenue growth
Nov 15, 2026Research agents (estimate)
Leadership Signals
CEO tenure, Founder-led
Oct 18, 2026Research agents (estimate)
Competitive Position
Category rank, Market share
Sep 24, 2026Research agents (estimate)

Why this Bluescore?

6 dimensions, each scored 0–10 against all 250 companies. Each adds its weight × score to the Bluescore — see methodology

Arena's Bluescore of 57.4 rests most on Talent & Research (8.8, top 4%) and Data & Model Moat (6.0, top 23%). Its weakest area is Investment Momentum (3.3, 25th percentile).

In plain English

Arena's Bluescore of 57.4 out of 100 is driven mostly by Talent & Research, where a score of 8.8 of 10 puts it at the 96th percentile among the AI250 and adds 14.2 of 16 points. The biggest drags are Investment Momentum at 3.3 of 10 (25th percentile) and Business Model & Revenue at 4.4 of 10 (45th percentile), which adds only 9.2 of 21 points. Business Model & Revenue would move the score most, because its 21% weight leaves 11.8 points unclaimed, the largest gap of any dimension.

Written by AI from the scores below and checked against them; updated nightly.

D1 · Business Model & Revenue (21%)
4.4
+9.2 of 21 pts · 45th pct
D2 · Product Depth (21%)
5.9
+12.4 of 21 pts · 73rd pct
D3 · Data & Model Moat (19%)
6.0
+11.5 of 19 pts · Top 23%
D4 · Talent & Research (16%)
8.8
+14.2 of 16 pts · Top 4%
D5 · Customer Adoption (13%)
5.2
+6.8 of 13 pts · 50th pct
D6 · Investment Momentum (10%)
3.3
+3.3 of 10 pts · 25th pct

Risks

Updated Nov 15, 2026

Studies such as The Leaderboard Illusion allege rankings favor big labs and are gameable
High
Revenue from model labs could compromise how neutral the public rankings appear
Medium
Crowd preference votes may not match enterprise accuracy, safety or domain needs
Medium
Young commercial company with a small team and limited enterprise track record
Medium
Faces competitors like Scale AI, Surge AI and Artificial Analysis
Low
Feedback