
tokenmixaiAudit of Veo 4 status in May 2026 — Google hasn't released it, but veo4free.io and other platforms are already charging subscriptions. Here's what's real and what's wrapper.
I went looking for Google Veo 4 last week. I had three browser tabs open with "Veo 4" landing pages, a credit card warmed up, and 30 minutes blocked to test the model that everyone on my feed seemed to be talking about.
Two hours later I was pretty sure I had not actually found Google's Veo 4 anywhere on the open internet. So I documented what I did find. Sharing here because if you're searching for "Veo 4" right now you're about to walk into the same situation.
As of May 12, 2026:
veo4free.io) has an unfilled template on its About pagedeepmind.google/models/veo is Google's official Veo product page. On May 12 it features Veo 3.1 prominently with the tagline "Video, meet audio. Our latest video generation model, designed to empower filmmakers and storytellers."
The page mentions Veo 3 and Veo 3.1. Not Veo 4. The DeepMind blog index has no Veo 4 entries. Google's AI/Gemini blog has no Veo 4 entries. The Google Cloud AI/ML blog has no Veo 4 entries. A direct request to blog.google/technology/ai/google-veo-4/ returns 404.
So whoever is selling "Veo 4" is not selling something from those URLs.
veo4free.io brands itself "Veo 4 — Free Multimodal AI Video Generator By Google DeepMind." Title tag, meta description, hero text all use the "by Google DeepMind" framing.
I fetched the site directly. Here's the relevant evidence:
The About page (literal text, copy-pasted from page):
Who We Are
[Company Name] is dedicated to [brief description of what your
company does]. Founded in [year], we have been [brief history
or achievement].
Our Mission
Our mission is to [your company's mission statement]. We believe
in [core values or principles] and are committed to [what you're
committed to delivering].
What We Do
We specialize in [your main services/products]:
[Service/Product 1] : [Brief description]
[Service/Product 2] : [Brief description]
[Service/Product 3] : [Brief description]
...
Contact Us
Email : [ your-email@company.com ]
Phone : [your-phone-number]
Address : [your-address]
That's the entire About page of a platform that's actively selling subscriptions and claiming Google DeepMind affiliation. The website builder template was never filled out.
The Blog page: "No blog posts. We are creating exciting content, please stay tuned!"
The pricing: $29.90 / $59.90 / $129.90 per month, credit-based generation, ~330 / 810 / 2040 videos per year per tier.
The model selector inside the generator UI: lets you pick between "Seedance AI," "Veo 4," "Seedance 2," "Veo 3.1," "Happyhorse," and "Nano Banana." Note that real Veo 3.1 appears alongside the unreleased "Veo 4" as separate selectable options.
I don't know what model handles a request when you select "Veo 4" on that platform. Could be Veo 3.1 routed through Google's API with a different label. Could be a different vendor's model entirely. Could be a future swap to real Veo 4 if and when Google releases the model. There's no documentation that tells you.
Artlist published a piece titled "Veo 4: What creators can realistically expect from the next generation of AI video," originally December 2, 2025, last updated April 20, 2026.
The article is candid about the status. Direct quote from their FAQ:
"Has Veo 4 been officially announced? No. Google DeepMind has not officially announced Veo 4. All current information comes from public research trajectories, industry reporting, and the evolution of previous Veo models."
"When is Veo 4 expected to be released? There is no confirmed release date. Based on Google's yearly update cycle and recent platform behavior, creators expect Veo 4 sometime in 2026, but this has not been officially confirmed."
So Artlist — a legitimate stock media + AI tools company that's a Google Veo partner — explicitly says Veo 4 is unreleased and they're publishing predictions, not features.
The predicted capabilities (their analysis, with industry confidence levels I'd assign):
These are educated industry expectations. They are not features you can rely on for production planning, and they are not features you can actually pay to access today.
If you map the Veo release cadence:
| Model | Release |
|---|---|
| Veo 1 | May 2024 (Google I/O) |
| Veo 2 | December 2024 |
| Veo 3 | May 2025 (Google I/O) |
| Veo 3.1 | Late 2025 / early 2026 mid-cycle refresh |
| Veo 4 | ? |
Google I/O 2026 is scheduled for late May. The previous two years' I/O have included a major Veo announcement. Best inference for the actual Veo 4 reveal is May 2026, within days of when I'm writing this. Could slip later. Won't be earlier.
This matters because it means today's "Veo 4" subscriptions are charging for something that doesn't exist for at least another week, possibly longer, possibly not until late 2026.
Four production-grade video generation models are shipping right now:
| Model | Max res | Audio | Max clip | Approx cost/sec | Best for |
|---|---|---|---|---|---|
| Veo 3.1 | 1080p | Native | ~1 min | $0.30–$0.75 | Storytelling with audio |
| Sora 2 | 1080p | Partial | ~20 sec | TBD | Cinematic shots |
| Wan 2.6 | 4K | None | ~10 sec | $0.01–$0.05 | Cost-sensitive 1080p volume |
| Kling O1 | 1080p | None | ~10 sec | $0.10–$0.25 | Stylized motion |
If you specifically want Veo capabilities, the three legitimate access paths today are:
1. Gemini app (gemini.google.com) - consumer subscription
2. Google Flow (flow.google) - creator-focused, credit packs
3. Vertex AI Veo API (cloud.google.com/vertex-ai) - developer, per-second pricing
That's it. Anything else is a wrapper layer.
If a platform fails 2 or more of these, it's not selling you Veo 4. It's selling you a wrapper that can swap its backend any time.
When I need video generation, I split workloads:
Personally I use TokenMix.ai for the LLM gateway part — 170+ language models behind one OpenAI-compatible endpoint, including Claude, GPT-5, Gemini 3.1 Pro, DeepSeek, Qwen, Kimi, GLM, MiniMax. The video models I still go to each provider's native API for, because video model coverage on aggregators is uneven and the per-second economics matter at scale. TokenMix's model intelligence tracker is also where I check Veo / Sora / Wan pricing changes month over month.
# typical setup for the LLM side
from openai import OpenAI
client = OpenAI(
api_key=os.environ["TOKENMIX_KEY"],
base_url="https://api.tokenmix.ai/v1",
)
# video generation: separate provider call to Vertex AI or Wan API
If you've tested any of the "Veo 4" platforms and got something specific about which model actually serves their requests, drop a comment — I'd genuinely like to know what's running under the hood there.
Full writeup with all pricing tables and the red-flag checklist on the main site at tokenmix.ai/blog/veo-4-reality-check-not-released-2026. All data verified as of May 12, 2026.