Artificial Analysis launches AA-Video-I2V v2.1, a benchmark for image-to-video models

Artificial Analysis launched AA-Video-I2V v2.1 and AA-Video-I2V-Silent v2.1, a benchmark for image-to-video models. Each prompt pairs a 1080p starting image with a prompt written for it and is tagged by use case, capability and visual style; models are judged at up to 1080p and ranked across 10 use cases and 10 capabilities. The overall ranking samples evenly across them. The results are Artificial Analysis's own.

The AA-Video-I2V v2.1 ranking rests on more than 58,000 votes from private US and UK evaluators over 1,000 prompts, and the Silent version on more than 44,000 votes over 500 prompts. Artificial Analysis says only 5% of the prompts are on its public arena, and that prompts rotate monthly to resist contamination.

Results it reports:

• Dreamina Seedance 2.5 ranks #1 overall, statistically tied with Seedance 2.0 and the MiniMax H3 family, and leads 11 category boards. At $34.12 per minute it is the most expensive model in the top 10.
• Dreamina Seedance 2.0 is #2, and #1 on Audio Synchronization and Lighting & Materials.
• MiniMax H3 (768p), an open-weights model, is #3 at $4.80 per minute, about a seventh of Seedance 2.5's price, and #1 on Text Rendering. It sits on the quality-versus-price Pareto frontier of both benchmarks.
• MiniMax H3 Max, post-trained by fal, is #4 with audio and #1 on the Silent version.
• Bach 1.0 Pro from Video Rebirth is #3 on the Silent version, statistically tied for #1, at $7.20 per minute.
• Rankings differ from text-to-video: Wan 3.0, which leads AA-Video-T2V v2.1, is #5 here, and Gemini Omni Flash 1.1 is #9 overall but #3 on Audio Synchronization.