Is Agnes Video 2.5 Worth It? Cheapest Top-10 Video AI

👁 114 reads

Agnes Video 2.5 costs $2.40 per minute of 1080p footage, making it the cheapest model in the top 10 of the Artificial Analysis text-to-video leaderboard. Generating 100 five-second clips costs $20, compared to $100 with Wan 3.0 and $284 with Seedance 2.5. However, MiniMax H3 sits 80 Elo points higher for $40 and, by my calculations, delivers more quality per dollar. You can find full details on the model in our Agnes Video 2.5 guide.

Text-to-Video Top 10: Price and Elo Ratings

All data comes from a single source: the AA-Video-T2V v2.0 leaderboard, recorded on October 11, 2026. Artificial Analysis tests each model in 1080p with audio and reports per-minute pricing under default settings. Utopai X (2nd place, 1149 Elo) has neither public pricing nor an open API, so it is omitted here.

Clip costs scale linearly: minute price / 60 x 5 s. A batch of 100 five-second clips equals 500 seconds, or 8 minutes and 20 seconds of video. Confidence intervals are rounded to the nearest whole point.

The full bar represents $284.33 (Seedance 2.5). The cost gap between the extremes is more than 14-fold.

Elo Points Per Dollar: Methodology

Price alone does not show how much visual quality you actually get. To evaluate this, I calculate a straightforward metric: how many Elo points above 1000 you get for each dollar of per-minute cost. On this leaderboard, 1000 represents Kling 3.0 1080p Pro, which AA uses as the baseline reference (confidence interval 1000-1000). The formula is: (Elo - 1000) / price per minute in USD.

A few caveats with this approach:

• Elo is non-linear. A 50-point difference at the top of the board is perceived differently than a 50-point gap in the middle.
• Results depend on the chosen baseline. A different reference point would shift the final ordering.
• AA pricing reflects a single standard configuration. Third-party providers are often cheaper (for instance, Agnes runs 33% cheaper in 1080p on crun.ai).

Agnes takes 3rd place, right behind the two MiniMax models. Calculating with crun.ai pricing ($1.61/min in 1080p) yields 35.4 points per dollar and 1st place, but that is a reseller discount rather than direct provider pricing.

What an 80-Point Gap Actually Means

Elo ratings translate directly into win probabilities in head-to-head comparisons. With an 80-point deficit (Agnes vs MiniMax H3), the lower-rated model wins approximately 39% of blind matchups; with a 99-point gap (vs Wan 3.0), that drops to around 36%. Agnes will not lose every comparison, but human evaluators favor the competitor roughly 6 out of 10 times. Meanwhile, Agnes sits just 17 points behind SkyReels V4 (9th place) with overlapping confidence intervals, putting both models at virtually identical quality. Yet SkyReels costs 8.75 times more.

When the Cost Savings Make Sense

• High-volume text drafts. When generating 10 iterations of a scene to pick one, spending $20 for 100 runs instead of $100 on Wan 3.0 makes a major difference.
• Storyboarding and animatics. Testing framing, camera movements, and pacing early, then rendering final output with a premium model.
• Short-form social content at scale. 9:16 vertical ratio with audio included costs $0.48 for the maximum 12-second clip in 1080p.
• Low-cost 2K rendering. At $3.30 per minute, 2K output is still cheaper than MiniMax H3 at $4.80.

When It Does Not Make Sense

• Image-to-video projects. On the image-to-video leaderboard, Agnes ranks 25th (1031 Elo). The leader, MiniMax H3 Max, scores 1195 and costs $4.80/min.
• Single hero assets. For one-off clips, the cost variance amounts to pocket change, while the drop in visual fidelity remains noticeable.
• Relying on the free Flash tier. Since October 10, users report that the free tier consistently errors out with 503 “video_queue_full” (GitHub issue #142, unaddressed by the company through Oct 11).
• Platform accessibility. Agnes is not hosted on fal.ai, Replicate, or OpenRouter. You need a dedicated account directly with the provider or a third-party aggregator.

The Verdict: Which Model to Pick

• Tightest budget, heavy text iteration: Agnes Video 2.5.
• Best overall balance: MiniMax H3 or H3 Max - double the price, significantly higher quality, and H3 Max also leads in image-to-video.
• Peak text-to-video quality: Wan 3.0.
• Google ecosystem integration: Gemini Omni Flash 1.1 at $9.12/min.

For per-second rates across additional models beyond the top 10, see our guide on how much a second of AI video costs.

Frequently Asked Questions

How much does a minute cost with Agnes Video 2.5?

Direct API pricing is $1.50 in 720p, $2.40 in 1080p, and $3.30 in 2K. Through crun.ai, rates drop to approximately $0.94, $1.61, and $2.14 respectively.

Is Agnes the cheapest model on the entire leaderboard?

Within the text-to-video top 10, yes. Lower-tier options like Boreal ($2.69/min, 838 Elo) cost roughly the same but deliver significantly worse output.

Is MiniMax H3 a better choice?

In most scenarios, yes: it delivers an 80-point Elo bump for double the cost. Agnes is only optimal when prompt iteration volume matters more than the fidelity of any individual generation.

// read moreAgnes Video 2.5: Cheapest Top-10 AI Video Generator, Pricing & GuideAI ToolsWhat Agnes Video 2.5 Can Do: 8 Prompt ExamplesAI ToolsOpen Video Models 2026: MAGI-2 vs Wan vs LTX-2.5 vs MiniMax H3Benchmarks