- Sora 2 vs Veo 3 is the wrong comparison now. Google shut Veo 3 down on 30 June 2026 and directs new work to Veo 3.1, so the live matchup is Sora 2 against Veo 3.1.
- On price it is not close at the top tier: Sora 2 is $0.10 per second of 720p, Veo 3.1 Standard is $0.40 — four times more.
- But tier beats vendor. Veo 3.1 Lite at $0.05 undercuts Sora 2, and Veo 3.1 Fast matches it exactly at $0.10.
- Sora 2 batch pricing halves to $0.05 per second, which is the cheapest frontier rate available for queued work.
- Veo 3.1 includes audio by default at every tier and reaches 4K; Sora 2 tops out at 1080p on the Pro tier.
Sora 2 vs Veo 3 is one of the most-searched comparisons in AI video, and it now carries a stale premise. Google has moved to Veo 3.1 and has since shut Veo 3 down, so a genuine 2026 decision is between Sora 2 and Veo 3.1. That correction changes the answer, because Veo 3.1 introduced tiers that Veo 3 did not have — and the tier you pick matters more than the vendor you pick.
Price: the comparison that actually decides it
| Model and tier | 720p | 1080p | 4K |
|---|---|---|---|
| Veo 3.1 Lite | $0.05 | $0.08 | — |
| Sora 2 | $0.10 | — | — |
| Veo 3.1 Fast | $0.10 | $0.12 | $0.30 |
| Sora 2 Pro | $0.30 | $0.70 | — |
| Veo 3.1 Standard | $0.40 | $0.40 | $0.60 |
All figures are per second of generated video, from the OpenAI and Google pricing pages on 18 August 2026. Sora 2 Pro also offers a 1024p step at $0.50.
Read that table twice, because the headline everyone repeats — “Sora is four times cheaper than Veo” — is only true against Veo 3.1 Standard. Against Veo 3.1 Fast the two are identical at $0.10, and against Veo 3.1 Lite, Google is half the price of Sora 2. Any comparison that names a vendor without naming a tier is not telling you anything useful.
The batch tier nobody mentions
OpenAI prices Sora 2 at $0.05 per second on the batch tier, and Sora 2 Pro at $0.15 / $0.25 / $0.35. Batch trades latency for cost: you submit the job and collect it later rather than waiting on a response.
For anything not interactive — overnight renders, bulk asset generation, a content pipeline that runs on a schedule — this makes Sora 2 the cheapest frontier-quality video model available, matching Veo 3.1 Lite on price while being a full-strength model rather than a reduced tier. If your workflow can tolerate the wait, this single setting is worth more than the vendor choice.
What each one gives you beyond price
| Sora 2 | Veo 3.1 | |
|---|---|---|
| Cheapest tier | $0.05 (batch, 720p) | $0.05 (Lite, 720p) |
| Top resolution | 1080p (Pro) | 4K (Fast and Standard) |
| Audio | Not priced separately in the API table | Included by default at every tier |
| Tiers available | 2 (standard, Pro) | 3 (Lite, Fast, Standard) |
| Batch discount | Yes, 50% | Not listed |
| Orientation | Portrait and landscape at 720p | Resolution tiers rather than orientation tiers |
Two differences matter more than the rest. Veo 3.1 reaches 4K and Sora 2 does not, which settles it if you need delivery-grade resolution without an upscale pass. And Veo 3.1 includes audio at every tier by default, which removes a separate sound step from the pipeline.
What a real project costs on each
| Scenario | Sora 2 | Veo 3.1 Fast | Veo 3.1 Standard |
|---|---|---|---|
| One 10-second shot | $1.00 | $1.00 | $4.00 |
| 10 shots x 10s (with 5 retries each) | $50.00 | $50.00 | $200.00 |
| A 60-second finished piece | $6.00 | $6.00 | $24.00 |
| 500 x 6s social clips | $300.00 | $300.00 | $1,200.00 |
The retry row is the honest one. A usable shot rarely lands first time, and at five attempts the gap between a $0.10 and a $0.40 model becomes $150 on a single small project. Model your budget on seconds of finished video, not seconds generated.
Which should you choose
Choose Sora 2 if your work is queued rather than interactive. The batch tier at $0.05 per second is the cheapest frontier rate on the table, and nothing Google lists matches a full-strength model at that price.
Choose Veo 3.1 Lite for high-volume, lower-stakes output — social clips, backgrounds, concept passes. At $0.05 per second interactive, it is half the price of Sora 2 for work that does not need the top tier.
Choose Veo 3.1 Standard only when you need 4K or the audio is doing real work. At $0.40 per second you are paying four times Sora 2, and that has to be justified by something specific rather than by a general sense that the flagship must be better.
Do not choose Veo 3. It is on a published shutdown schedule. If it is already in production, plan the migration to Veo 3.1 rather than treating it as a live option.
The full per-second table for every model we track, including Kling, Wan and the image generators, is in our AI video generation pricing guide. For text-model budgets the AI API cost calculator does the same job, and the AI models database holds the specifications behind both.
Frequently asked questions
Is Sora 2 cheaper than Veo 3?
Against Veo 3.1 Standard, yes: Sora 2 is $0.10 per second versus $0.40, four times cheaper. Against Veo 3.1 Fast they are identical at $0.10, and against Veo 3.1 Lite at $0.05 Google is half the price. The tier decides it, not the vendor.
Is Veo 3 still worth using in 2026?
No. Google shut Veo 3 down on 30 June 2026 and points new work to Veo 3.1. Existing Veo 3 deployments need a migration plan rather than a pricing comparison.
Which produces higher resolution, Sora 2 or Veo 3.1?
Veo 3.1. Its Fast and Standard tiers generate 4K at $0.30 and $0.60 per second. Sora 2 Pro tops out at 1080p, at $0.70 per second.
Does Sora 2 or Veo 3.1 include audio?
Google prices Veo 3.1 with audio included by default across Lite, Fast and Standard. The OpenAI API pricing table for Sora 2 lists rates per second of video without a separate audio line.
What is the cheapest way to generate AI video?
Sora 2 on the batch tier and Veo 3.1 Lite both cost $0.05 per second of 720p video. Batch requires that your workflow can wait for the result; Lite is available interactively but is a reduced tier rather than a full-strength model.
Prices are the published list rates from OpenAI and Google, checked on 18 August 2026. Both providers change pricing and deprecate models without notice, so confirm against the vendor page before committing a budget.
