Guides
Veo 4, Gemini Omni Flash, "Wan 3.0 Open Weights": A Checklist for Reading Model Claims
ยท RenderBob team
Model names, version numbers, and release claims are getting hard to track. Several recent items conflict, and a few rumours have been repeated as fact.

Model names, version numbers, and release claims are getting hard to track. Several recent items conflict, and a few rumours have been repeated as fact.
Examples from this month
- Veo 4 and Gemini Omni Flash. Coverage uses the names in ways that suggest they may overlap or be confused. Confirm on Google's own pages which product, tier, and interface you are looking at.
- Wan 3.0 open weights. Sources disagree on whether weights are released, partly released, or promised. Until you can download them from the official repository, treat open weights as unverified.
- A Seedance 2.1 rumour. The name circulated without a primary source. We do not treat it as a product.
Is there a primary source?
An official page, repository, or changelog beats a recap article.
What is the date?
Many pages are republished or updated. Check when the claim was first made.
Which tier?
Flash, standard, and pro tiers often have different limits.
Is it available to you?
Region, plan, and waitlists matter.
Has anyone reproduced it?
Look for tests on material that was not chosen by the vendor.
Aggregator pages
Roundup pages and quarterly recap sites are often written quickly and sometimes contain errors. We discarded one such recap while preparing this week's articles. Use them to find leads, and confirm the facts elsewhere.
What to watch
Vendors tidying up their naming, and community trackers that log model versions with sources.
More from the blog
- "$300 Million Versus $70 Million": How to Read Budget Claims for AI-Made Films
The producers of Bitcoin: Killing Satoshi say a $70 million AI-assisted film would have cost over $300 million to shoot conventionally. Both numbers are theirs, and the $70 million was not itemised.
- Text in Images Is Getting Reliable, but Proofread It Anyway: Ideogram 4.5, GPT-Image-2.5 and Nano Banana 2.1
Ideogram 4.5, GPT-Image-2.5, and Nano Banana 2.1 advertise reliable text in images. Ideogram's own guide still shows wrong numbers, extra headings, and lines that wrap and drop punctuation.