The LMArena startup, which evolved from a UC Berkeley research project, announced that it has raised $150 million in its Series A round at a valuation of $1.7 billion. Interestingly, just seven months ago, the company secured $100 million during its seed stage with a valuation of $600 million. This brings the total funds raised over a brief period to $250 million. LMArena is renowned for its crowdsourced performance ratings of AI models. Users of the site input a text query, which is sent to two models. The user then selects which model performed better. These results, gathered from over 5 million users monthly across 150 countries and 60 million dialogues, form the basis for the rankings. The platform evaluates models across various tasks, including text, web development, vision, text-to-image conversion, and other criteria.
Among the models tested are various versions of OpenAI GPT, Google Gemini, Anthropic Claude, and Grok, as well as specialized models for image generation, text-to-image transformation, or reasoning. The company began as Chatbot Arena, an open research project created by UC Berkeley researchers Anastasios Angelopoulos and Wei-Lin Chang, initially funded through grants and donations.
LMArena’s ratings have become a popular tool for AI model developers. In April, a group of competitors published an article claiming that LMArena’s partnership with companies like OpenAI, Google, and Anthropic allowed them to manipulate benchmark results. LMArena categorically denies these accusations.
In September, the company launched a commercial service called AI Evaluations, offering enterprises, labs, and developers the opportunity to assess models through crowdsourcing. By December, LMArena’s annual revenue reached $30 million, just four months after launching the service.
