LMArena, an AI model evaluation startup that began as a UC Berkeley research project, announced Tuesday it raised $150 million in Series A funding at a $1.7 billion post-money valuation. The round, led by Felicis and UC Investments, represents a remarkable trajectory for a company that launched its first commercial product just four months ago in September 2025.
The funding round included participation from Andreessen Horowitz, The House Fund, LDVP, Kleiner Perkins, Lightspeed Venture Partners, and Laude Ventures. Many of these firms also backed LMArena's $100 million seed round in May, which valued the company at $600 million. The new valuation represents a tripling in value over just seven months.
LMArena's rapid ascent stems from its crowdsourced approach to AI model benchmarking. The startup operates a consumer-facing website where users input prompts that get sent to two anonymous AI models simultaneously. Users then select which model provided the better response, creating millions of data points that fuel comprehensive performance leaderboards.
The platform now attracts more than 5 million monthly users across 150 countries who have collectively generated over 60 million conversations. These crowdsourced evaluations rank various AI models across multiple capabilities including text generation, web development, computer vision, text-to-image generation, and other specialized tasks.
In September, LMArena launched AI Evaluations, its first commercial service targeting enterprises, model laboratories, and AI developers. The offering allows companies to hire LMArena to perform systematic model evaluations using its massive user community. Customers receive detailed evaluation results plus access to underlying feedback data samples for verification purposes.
The commercial service gained immediate traction. LMArena disclosed that AI Evaluations reached a $30 million annualized consumption run rate as of December, representing significant revenue generation within the first four months of availability. This consumption-based pricing model differs from traditional software-as-a-service subscriptions, with customers paying based on actual evaluation usage.
LMArena's benchmarking platform has become critical infrastructure in the generative AI ecosystem. As companies invest billions developing large language models, objective performance measurement becomes essential for guiding development priorities and validating improvements. Traditional AI benchmarks use predefined prompts with correct answers, but LMArena's approach captures real-world user preferences across diverse use cases.
The company plans to deploy the new capital across three primary areas: covering operational costs for running its high-traffic platform, financing AI research initiatives, and expanding its engineering team. The platform's computational requirements are substantial given it processes millions of model queries monthly across multiple AI systems.
Founded in 2023 by two UC Berkeley researchers, LMArena originally operated as an academic project providing free model comparisons. The transition to commercial operations came as enterprises recognized the value of large-scale, unbiased model evaluations that reflect actual user preferences rather than narrow benchmark performance.
The $1.7 billion valuation four months after product launch highlights investor conviction that AI model evaluation represents a sustainable business category as the AI industry matures and companies demand rigorous performance verification before deployment decisions.
