HeyGen's first general video model lands near the top of a blind-test leaderboard

HeyGen, best known for AI avatars, now has a general-purpose video model. On Artificial Analysis’ text-to-video leaderboard, which scores models from blind human votes, HeyGen Video 1 sat third on October 25 with an Elo of 1,160, behind Wan 3.0 (1,175) and Utopai X (1,167). The margins fall inside each model’s confidence range, so the top group is effectively a tie. HeyGen’s model page goes further and calls it “the #1 video model”. A Dealroom headline credits it with first place on the board’s silent, no-audio ranking, which we could not check.

HeyGen says the model is post-trained on MiniMax H3. It makes 5 to 15 second clips with sound from a prompt, an image or a reference video. The help center lists $0.03 per second at 768p, or $0.015 during a 50% launch offer that runs to the end of October. Artificial Analysis lists $2.40 a minute, against $12 for Wan 3.0 and about $34 for Seedance 2.5.

If you already pay for HeyGen, you can try it in Creative Studio’s model menu. If you only need short generative clips, it is worth testing against Kling or Veo on price per usable shot.