Slop TVNewsLatest
News

Pruna's P-Video-2-Pro lands 11th on the AI video leaderboard

P-Video-2-Pro is built on MiniMax H3's open weights, and H3 itself rose to third on Artificial Analysis's text-to-video board.

By

·2 min read

Illustration: Pruna's P-Video-2-Pro lands 11th on the AI video leaderboard
Illustration: SLOP TV News. Logos are trademarks of their owners.

Key takeaways

  • Pruna's P-Video-2-Pro entered Artificial Analysis's text-to-video leaderboard at an Elo of 1125 in eleventh place, on the board as read on October 31, 2026.
  • MiniMax H3 at 768p, the model P-Video-2-Pro is built on, moved up to third on the same board at 1161 Elo.
  • Pika's H3 Flex, released on 29 October, and Utopai Studios' Utopai X are also post-trained on MiniMax H3; Pika says on its blog that H3 Flex ranks first among open-weight models on DesignArena's text-to-video board, and Utopai X ranks second on Artificial Analysis's image-to-video board.
  • P-Video-2-Pro generates 5 to 15 second clips at 480p or 768p from text or a first and last frame, with generated audio, from $0.01 per second.

Pruna's P-Video-2-Pro entered Artificial Analysis's text-to-video leaderboard at an Elo of 1125 in eleventh place. The Japanese daily Synon reported the placement in its 31 October roundup of video-generation news, which gives the board as read that day. The model is built on MiniMax H3, and H3 itself moved up to third on the same board at 1161 Elo. Wan 3.0 leads at 1175 and Utopai X is second at 1166.

Positions on the board are a snapshot. Artificial Analysis fits its Elo ratings from pairwise human-preference votes, and a model's place can shift as new votes arrive. Artificial Analysis lists P-Video-2-Pro among six models added to the board in the last 30 days.

Pruna describes P-Video-2-Pro as based on MiniMax H3, and sells it in three generation modes: Cost, Speed and Quality. Pruna's API documentation gives clips of 5 to 15 seconds at 24 frames per second, at 480p or 768p, with 1080p, 2K and 4K available by upscaling the 768p video. The model takes a text prompt or a first and last frame, its output includes generated audio, and it does not accept audio as an input. List prices run from $0.01 per second for 480p in Cost mode to $0.075 per second for 768p in Quality mode.

MiniMax H3 is the base several makers are now building on. MiniMax's own repository says it releases "the complete model weights to support further development, including fine-tuning," under the MiniMax H3 Community License Agreement. Pika released H3 Flex on 29 October, post-trained on MiniMax H3, and says on its blog that the model ranks first among open-weight models and seventh overall on DesignArena's text-to-video leaderboard. Utopai Studios says its Utopai X, also post-trained on H3, took second place on Artificial Analysis's image-to-video board in the 29 October results.

For a creator, the entry means H3-class generation is now available in several versions and through several storefronts. Wiro lists P-Video-2-Pro with Speed and Quality modes, while Pruna's own API also lists a cheaper Cost mode.

P-Video-2-Pro runs through Pruna's API, or through the resellers that list it, including each::labs and Wiro.

Sources

  1. pruna.ai - Pruna's own model page: built on MiniMax H3, the Cost, Speed and Quality modes, prices from $0.01 per second, 768p, generated audio
  2. docs.api.pruna.ai - Pruna's own API docs: 480p and 768p native with 1080p, 2K and 4K upscales, 5 to 15 second clips, 24 fps, text-to-video and first/last-frame modes, per-second pricing
  3. github.com - MiniMax's own repository: complete H3 model weights released for further development including fine-tuning, under the MiniMax H3 Community License Agreement
  4. synon.co.jp - the outlet's 31 October read of the Artificial Analysis text-to-video board: P-Video-2-Pro at 1125 in eleventh, MiniMax H3 768p at 1161 in third
  5. pika.art - Pika's own blog: H3 Flex post-trained on MiniMax H3, released 29 October, ranks first among open-weight models and seventh overall on DesignArena's text-to-video board
  6. utopaistudios.com - Utopai's own blog: Utopai X post-trained on MiniMax H3, No. 2 on Artificial Analysis's image-to-video board in the October 29, 2026 results
  7. artificialanalysis.ai - the leaderboard itself: top of the board (Wan 3.0 1175, Utopai X 1166, MiniMax H3 768p 1161), Elo from pairwise human-preference votes, P-Video-2-Pro among six models added in the last 30 days
  8. wiro.ai - reseller listing: Speed and Quality modes, 480p or 768p at 24 fps with synced audio, first and last frame conditioning
  9. eachlabs.ai - reseller listing: per-second pricing by resolution and mode, 5 to 15 second duration