Two PhDs' Arena Tops $100M; Leaderboard to Read
Chapters
Eu 36kr reports Arena's valuation climb
Eu 36kr reported on October 15, 2026, that Arena, a business that ranks AI models, has seen its valuation rise 80% in nine months. The outlet's headline gave the valuation as 20.7 billion yuan. Eu 36kr is the only outlet in this coverage carrying the valuation and revenue figures; they have not been confirmed elsewhere.
Arena designs questions, administers exams and scores AI models including ChatGPT and Claude, Eu 36kr said. The outlet added that AI evaluation, once the work of university laboratories, has become a hot business in the capital market.
Eu 36kr described Arena as one more AI business on a fast climb. It called the speed of the company's revenue growth even more striking than the valuation gain and said similar stories keep emerging.
The short version
Eu 36kr reported that Arena, an AI model ranking business that began as a Berkeley research project, saw its valuation rise 80% in nine months. Arena disclosed annualized revenue above $100 million, Eu 36kr said. The story gives small and mid-sized business owners an AI leaderboard and its financial backing to weigh.
- Chatbot Arena was started in 2023 by UC Berkeley PhD candidates Jiang Weilin and Anastasios Angelopoulos, Eu 36kr said.
- Ordinary users vote between two anonymous model answers, and the votes produce the rankings, according to Eu 36kr.
- The 20.7 billion yuan valuation comes from Eu 36kr's headline; the report has not been confirmed elsewhere.
Two PhDs launch AI model rankings at Berkeley
The project began in 2023, when Jiang Weilin and Anastasios Angelopoulos, two PhD candidates at UC Berkeley, launched Chatbot Arena, Eu 36kr said. Their aim was to answer which model is actually smarter among ChatGPT, Claude, Gemini and similar models.
At the time, model makers presented test results in mathematics, programming and knowledge Q&A with every new release, Eu 36kr said. A high score in those tests does not mean users find the model easy to use, the outlet said. The Berkeley team came up with an idea: let ordinary people act as examiners.
On Chatbot Arena, users pick the better of two anonymous model answers, and the votes build a ranking list, Eu 36kr said. Voters do not know which answer comes from OpenAI and which from Anthropic.
Model makers court the Arena leaderboard
A large model company that releases a new product now hopes to rank high on the Arena leaderboard, alongside showing its own test results, according to Eu 36kr. The outlet said a research project has gradually become an important evaluation platform in the AI industry.
The team set up the commercial company Arena Intelligence in 2025 and began to accept venture capital, Eu 36kr reported.
Arena disclosed in June 2026 that annualized revenue had passed $100 million, about 670 million RMB, eight months after commercial launch, Eu 36kr said. The report has not been confirmed elsewhere.
| Date | Event reported by Eu 36kr |
|---|---|
| 2023 | Jiang Weilin and Anastasios Angelopoulos launch Chatbot Arena at UC Berkeley |
| 2025 | The team establishes the commercial company Arena Intelligence |
| June 2026 | Arena discloses its annualized revenue figure |
| October 15, 2026 | Eu 36kr publishes its report on Arena |
Tron's take
My reading is that the news gives owners a scoreboard to understand, not a purchase signal. Eu 36kr said Arena's rankings come from ordinary users voting on anonymous answers. A high rank therefore describes what voters preferred in open-ended chats. Whether a model handles a company's invoices, tickets or contracts is a separate question. That is my reading of the news, not a reported result.
I would treat the valuation and revenue figures as unconfirmed, since they rest on one outlet's report. I would also not read a company's valuation as proof that its rankings are accurate.
On timing, I would pick a model that already works on a few real business tasks and use the leaderboard only to build a shortlist when a new release arrives. Model makers will keep shipping versions, and rankings will keep moving. XL.net's earlier report, Kimi K3 Highlights Benchmark Limits, covers the limits of leaderboards.
Questions I'd expect
What does Arena do?
Eu 36kr said Arena designs questions, administers exams and scores AI models including ChatGPT and Claude. Ordinary users vote between two anonymous answers, and the votes form a ranking list.
Who started Arena?
Eu 36kr said Jiang Weilin and Anastasios Angelopoulos, two PhD candidates at UC Berkeley, launched Chatbot Arena in 2023. The commercial company Arena Intelligence followed in 2025.
Have the valuation and revenue figures been confirmed?
The 20.7 billion yuan valuation and the revenue figures come from Eu 36kr's report; they have not been confirmed elsewhere.