A film studio, not an AI lab, now sits near the top of the world's video-generation rankings. Utopai X, the in-house video model of California-based Utopai Studios, has reached No. 2 globally on Artificial Analysis' Text-to-Video Leaderboard With Audio, with an Elo score of 1,150 as of the September 30 update - the highest position held by any model from a US-based company on the board.
The leaderboard is built on blind human-preference testing: viewers compare two clips generated from the same prompt without knowing which model made which, and the votes produce Elo scores. On that board, Utopai X trails only Alibaba's Wan 3.0 (Elo 1,157) and sits ahead of ByteDance's Dreamina Seedance 2.5 (1,143) - as well as the MiniMax H3 (768p) base model it was built on, which scores 1,138. Artificial Analysis confirmed the model is a post-train of MiniMax H3, meaning Utopai out-ranked both its own foundation model and MiniMax's in-house H3 Max variant.
The margins deserve a caveat. Utopai X's confidence interval (1,139-1,161) overlaps with Wan 3.0's point estimate, and Artificial Analysis itself labels the rank as a range spanning roughly No. 1-3 rather than a fixed position. The company's claimed strengths in audio sync, physics, lighting and camera control are development targets reported by the vendor, not independent benchmark categories. And unlike its ranked rivals, Utopai X has no public API and no per-second pricing - it exists in exactly one place, inside Utopai's own production platform.
That platform is the real product. PAI, Utopai's production-intelligence system, breaks a screenplay into scenes and shots while keeping characters, locations, props and visual references attached to the project. AI "production assistants" help filmmakers plan shots and iterate versions inside the same workspace, and selected footage moves into editing timelines and established post tools like Premiere and DaVinci Resolve. Utopai argues the flywheel - artists' feedback from real productions feeding the model - is what a research lab cannot replicate.
The company behind it is an unusual entrant. Founded in 2022 as Cybever and renamed Utopai Studios in 2025, it is led by co-founders Cecilia Shen (CEO, previously in data science and AI solutions at Google X) and Jie Yang (CTO), with Zijian He as chief scientific officer. Forbes estimated its valuation at $1 billion in May 2026, and Variety has described it as the world's largest independent AI-native film and television studio, with 25 projects in development.
The 2027 slate is the actual test of the ranking. It includes three films - animated feature "The Most Serious Fart" (from Mike Bender's 2023 children's book), family film "Half Moon" (in production in Germany since August), and the long-gestating historical epic "Cortes," written by Oscar nominee Nicholas Kazan - plus two series: "Space Nation," reportedly the first AI production sold to traditional buyers including Brazil's Globoplay at non-AI market prices, and "Journey to the West: The Lost Five Hundred Years," developed with China's Huace Film & TV. Utopai says potential combined presales for "Space Nation" and "Cortes" could reach $110 million, conditional on delivery. NBA stars Carmelo Anthony and James Harden are among its investors and partners.
The model's rise lands in a sensitive moment for Hollywood. The 2023 SAG-AFTRA and WGA agreements set consent and pay rules for digital replicas and bar studios from forcing writers onto AI. Shen has said Utopai trains on non-infringing data, works with the unions and keeps creative decisions with humans. Whether an Elo score translates into films audiences actually sit through will be settled in theaters in 2027 - a leaderboard measures clips, not features.
Comments (0)
Log in to join the discussion
Log InNo comments yet