HyperFrames Partners with DeepMind and Kaggle to Launch Code2Video Bench
On September 22, Kaggle and HeyGen's HyperFrames team, together with Google DeepMind, launched Code2Video Bench, positioned as the first benchmark for agentic video tasks, measuring models' ability to drive video generation with code.
Confirmed
- The benchmark was built by the HyperFrames team (under HeyGen) in collaboration with Google DeepMind and Kaggle to evaluate code-driven video generation.
- According to Kaggle's Megan Risdal, the benchmark measures whether frontier models can generate deterministic, high-quality motion graphics code from natural-language creative briefs.
- Multiple reposters (nikolamr64990, AidenTechAi, aftahiai) consistently relayed the publishers' view: the agentic video stack is being built on code generation, but existing code benchmarks like SWE-bench do not equate to code-driven video capability, hence the need for a dedicated benchmark.
Why it matters
- The benchmark fills the gap of standardized evaluation for agentic video, providing a unified yardstick for the emerging capability of driving video generation with code; Kaggle's involvement also suggests a potentially community-open evaluation and competition ecosystem.
2026-09-22 ~ 2026-09-22 · 6 related posts
Primary sources
- HyperFrames and Google DeepMind Launch Code2Video Bench for Agentic Video Tasks — aftahi_ai · 2026-09-22
- Code2Video Bench launched with Google DeepMind and Kaggle for agentic video evals — Aiden_Tech_Ai · 2026-09-22
- [source] HeyGen and Kaggle launch Code2Video Bench for motion graphics code generation — MeganRisdal · 2026-09-22
3 near-duplicate retellings: Aiden_Tech_Ai · nikola_mr64990 · nikola_mr64990