About Rewind.ai
400+ AI tools, 400+ models and 100+ languages, all in one place. No credit card required to start.
You should be able to run AI tools and see real results before you decide to pay for anything. That is the whole premise. We built the infrastructure to make it possible at no cost to you.
Why We Built This
GPU compute is expensive, and most AI products cover that cost with $20-50/month subscriptions you have to commit to upfront. That is a hard ask when you have not yet confirmed the tool fits your workflow.
We run some open-source models, such as FLUX.2 Klein, SDXL and Wan 2.2, on our own GPUs, which keeps them cheap enough for a free daily allowance. Other models, including chat models, run through their providers.
Premium models such as GPT, Claude and Gemini carry real API costs. They run on tokens from a token pack or a plan.
How It Works
- Free tokens: every account starts with 10,000 tokens; 5,000 more are added each day. No card needed.
- Self-hosted models: FLUX.2 Klein, SDXL, CogVideoX and Wan 2.2 run on our own GPUs at the lowest token cost.
- 400+ external models: GPT-4, Claude, Gemini and hundreds of others from major providers, each billed per token from the same balance.
- One token system: a single balance covers every tool; the token cost for each request is displayed before you submit.
How We Operate
Open Source Models
Every model we host carries an MIT or Apache 2.0 license. Outputs belong to you for personal or commercial use.
Clear Pricing
Token cost is shown before each request, and purchased tokens cover every model.
Privacy
We do not sell your data and do not use your inputs to train models. Requests processed by self-hosted models never leave our servers.
Honest Metrics
Numbers on the site reflect actual usage. If a feature is not ready, we say so rather than list it as available.
Tech Stack
The frontend is Next.js; the backend runs on Django 6.x. Model inference is handled by a FastAPI service on NVIDIA A100 GPUs hosted at Vultr. PostgreSQL manages persistent data, Redis handles caching and Nginx routes traffic. Self-hosted models run under their respective open licenses (Apache 2.0, MIT, OpenRAIL).
Looking for the old Rewind?
This site is not affiliated with the original Rewind AI (the Mac app and Pendant, later Limitless). That product shut down in December 2025 after Meta acquired Limitless — read the full story on what happened to Rewind AI.
Contact
Questions or issues? Email us at [email protected] or use the contact page.