October 9th, local time 8thAI MODEL EVALUATION PLATFORM Arena THE ARENA ANNOUNCED THE COMPLETION OF $200 MILLION (THE CURRENT EXCHANGE RATE IS ABOUT RMB 13,433 MILLION) ROUND B OF FINANCING, WHICH IS NOW BEING FINANCED BY THE GOVERNMENT OF THE DEMOCRATIC REPUBLIC OF THE CONGOValue reached $3.1 billion(The current exchange rate is about RMB 20,817 million). Arena was originally a research project at the University of California at Berkeley in 2023, where the AI model was ranked by popular vote。
Arena previously disclosed that the annualization of this June amounted to $100 million (the current exchange rate is approximately RMB 672 million)。
The current round of financing is co-financed by light-speed investments and by institutions such as Khosla Ventures, Salesforce Ventures, 01 Advisors, Dale Science and Technology Capital, Endeavoor Catalest, a16z and Felicis。
In January this year, Arena completed a round of financing of $150 million (current exchange rate approximately RMB 1,007 million)Post valuation $1.7 billion(The current exchange rate is approximately RMB 114.1.6 billion) and the annualizations received US$ 30 million (the current exchange rate is about RMB 201 million). In a short period of about 10 months, Arena's valuation was nearly doubled。
Arena 's Assessment PlatformFree access to individual usersI don't know. The user can enter a hint, or allow AI to prepare a program, develop a project, and compare and rate the results of the different models. According to Arena, the platform attracts tens of millions of visitors each month。
Last September, Arena launched the AI Models Development Agency and Business Services, which provides detailed model performance analysis using feedback from community users。
THIS YEAR, AI, R & D FOUND THAT THE FLAG MODEL COULD PASSRules for matching baseline testsHigh marks are not necessarily matched by real capacity. Enterprises are no longer satisfied with standardized test results and would like to know which model is better suited to their business needs。
In the financing announcement, Arena states: “AI has grown faster than we can assess. Once the model finds itself tested, a fixed benchmark test will lapse. The world needs a neutral third party to test whether AI is safe in its actual use and is acting in accordance with human expectations. Arena began to assume that role."
For that reason, Arena is on the listAdditional alignment assessmentThe focus is on whether the model will perform unauthorized operations not required by the user, whether it will be crowned, attribute words or facts to the wrong source, and whether it will be falsely described as having fulfilled its mission. Arena describes the last act as “deceptive completion”。
On the initial alignment list, the OpenAI multi-mode model ranked at the top, Claude Opus 5.5 sixth, Claude Fable ninth。