The Network for Open Intelligence. Host models, run secure inference, and deploy agents verifiably onchain. Backed by @a16zcrypto, @cbventures, and more.
Introducing OpenGradient Chat, a privacy-first generative AI platform.
Every frontier model in one app. We can't read what you type, and neither can anyone else.
Finally, an AI you can tell anything.
(1/n)
It sits right next to ChatGPT, Claude, Gemini, Seed, and Grok, so you can switch models with one tap.
Go ask the thing you actually wanted to ask. (∇, ∇)
https://t.co/GA4YzjTsdK
3/3
Uncensored AI chat is now live on OpenGradient Chat
Hermes 4 405B by @NousResearch is an open model that answers straight.
Ask anything you want, privately. (∇, ∇)
1/3 🧵
The model won't refuse. It won't lecture about why it can't help you.
Hermes was built on one idea: the AI works for you, not for some content policy.
And it's private by default.
We don't mine your prompts or sell them, and they never get used to train anyone's model.
2/3
Image Studio is now live on https://t.co/GA4YzjTsdK
Generate images across Gemini, ByteDance, and xAI models. Private by default.
More models coming. (∇, ∇)
Sample generated image - using Nano Banana 2
PROMPT:
Mirror-world infinity portraiture - a Midjourney Prompt: [SUBJECT] walking in profile along a razor-thin horizon line on a [SURFACE] with a perfect mirror reflection below forming a symmetrical diamond geometry, shot at eye-level with [CAMERA_SYSTEM] and [FOCAL_LENGTH] lens on [FILM_STOCK], ultra-minimalist fine-art editorial style, subject rendered in [CONTRAST_TREATMENT] against a vast unbroken field of [COLOR1] sky occupying the upper two-thirds of the frame, flawless still-water reflection below in [COLOR2] mirroring the full figure inverted, absolute horizon line bisecting the composition with surgical precision, zero background detail, infinite negative space [GARMENT_DETAIL] catching ambient light, Hasselblad medium format tonal rendering, gallery exhibition quality --v 8.1, DO NOT ADD "AI generated" text.
5/ Bring the best model to the things you'd never paste into someone else's logs.
Finally, an AI you can tell anything.
Fable 5. Live now in OpenGradient Chat. ↓
https://t.co/ArIP2gijy0
Fable 5 is now live in OpenGradient Chat.
Anthropic's newest model, running where the conversation has no audience. We can't read what you type, and neither can anyone else.
The frontier model and the private place to use it are now the same place.
(1/n)
4/ Capability was never the hard part of private AI. Trust was.
Everyone says "private." Most ship your words through plaintext and a chain of middlemen who promise not to look. That is branding.
OpenGradient Chat makes privacy a property of the system, not a line in a policy.
This looks like a toy until you build it into a real flow. One thing to know: it's CC BY-NC-SA licensed, so keep it to research, demos, and non-commercial projects.
Try IDM-VTON on the hub: https://t.co/CMM3LVe1kr
OpenGradient Model Highlight: IDM-VTON
Upload a photo of a person and a photo of any garment. The model returns that exact person wearing that exact garment, photorealistic, posed correctly, looking like a picture they actually took.
Diffusion-based virtual try-on, live on our model hub. 🧵👇🏻
Picture a shopping agent that puts the item on a person from one photo, or a styling app that builds a lookbook from a single selfie. Two images in, a finished shot out.
6/
Introducing OpenGradient Chat, a privacy-first generative AI platform.
Every frontier model in one app. We can't read what you type, and neither can anyone else.
Finally, an AI you can tell anything.
(1/n)