OpenGradient

How to Direct AI Films with Seedance 2.5

Part 1: Continuity

New models just landed on https://dreamfort.ai

> GPT-6 Luna: OpenAI's fastest GPT-6, built for everyday work.
> Sonnet 5.5: near-frontier quality for writing, code, and analysis.
> Wan 3.0 video now has a cheaper 480p option. Same 30-second takes, fewer credits.

All private!

Private Super Intelligence >

Ask like no one's watching.

https://dreamfort.ai

Check everything here https://portal.opengradient.ai/inference-usage

Our biggest week yet!

OpenGradient averaged 14,000+ private inferences a day and processed over 2 billion tokens.

That brings us to 842K+ inferences and 8.7 billion+ tokens, and we couldn't read a single prompt.

This is a reminder that every prompt is encrypted on your device

Claude Opus 5.5 is now live on DreamFort.

Anthropic's newest Opus model, built for serious coding and AI agents, and their best Opus yet for reading documents, charts and screenshots.

Use it privately on Dreamfort: https://dreamfort.ai/agent

Try Dreamfort today: https://www.dreamfort.ai/?utm_source=x&utm_medium=post&utm_campaign=aivideo&utm_content=0925-5vh3

Just use Dreamfort.

AI labs agreed on safety for about 4 seconds 💀

→ They shake hands on safe superintelligence
→ But Mr. President said, "Whoever wins AI, wins!"

Which lab would you trust with your prompts? 👀

Use Private AI → https://dreamfort.ai

Wan 3.0 Prime is now live on https://dreamfort.ai/video

It's the highest-fidelity tier of Wan 3.0.

Turn text or an image into a clip up to 30 seconds long, with audio and reference images for more control.

Private and uncensored by default.

Machine-frequency payments need low-cost settlement and fast finality. Base provides both, which is why these rails run here.

If you're building agents on Base that need to pay for AI compute programmatically, the rails are already live.

Start building: https://docs.opengradient.ai

All of it is public.

Settlement activity: https://basescan.org/address/0xA71c98C0d28c63DBf7ca36f878d6e8De6Bd213Bb
Network usage: https://portal.opengradient.ai/inference-usage

These transactions come from live product usage.

https://Dreamfort.ai, the private AI app built on OpenGradient, drives 57% of all inference on the network. That usage settles through these same x402 rails on Base.

Each settlement transaction batches multiple inferences underneath.

180K transactions on Base now represent 660K+ inferences served across the network, all of them run privately inside secure enclaves.

OpenGradient x402 settlement on @base

AI inference on OpenGradient settles through x402, with machine-native payments batched onchain.

180K settlement transactions on Base since launching in April.

742,000 AI conversations served. 6.6 billion tokens processed.

All encrypted, so nobody can read them. Not hackers, not employees, not even us.

That's what private AI means.

https://portal.opengradient.ai/inference-usage

but... you better go private. 👀

https://www.dreamfort.ai/

"the end is coming!"

ai could kill humans in less than a decade!

nothing to worry about here...

We asked GPT Image 2.5 to make its own launch poster. This is what it handed back.

One prompt, every letter on the sign correct; then we dropped the logo on.

GPT Image 2.5 is now on https://Dreamfort.ai: text that spells, edits that change only what you point at, and every