On this page
If you want to use open models like Qwen and DeepSeek through an API, you usually pay per token, and the bill moves every month. Off Grid AI Boost gives you one API key for open models on our own GPUs, at one flat monthly price, and we never store or train on your prompts. It’s open for pre-order now at half price for the first cohort, and API keys go out in January 2027.
What is Off Grid AI Boost?
Boost is hosted inference for open-weight models, run by the team behind Off Grid AI. You get one API key that works with every model on Boost, including Qwen 3.8 and DeepSeek V4.1 Flash. It covers text, vision, speech, voice, embeddings, images and video. Usage limits for each plan haven’t been published yet.
You pay a flat price, monthly or yearly, so there’s no per-token maths and no surprise bill at the end of the month.
Why does a flat price matter?
Per-token pricing is cheap when you use a little and hard to predict when you use a lot. An agent that loops, a batch job or a busy app can multiply a bill overnight.
With a flat price you know the cost before the month starts, so you can build, test and run things without watching a meter.
What does it cost?
The first cohort pays half price, and keeps that price for as long as they stay.
| Regular price | Cohort 1 (50% off) | Cohort 2 (25% off) |
|---|---|---|
| $9.99 a month | $4.99 a month | $7.49 a month |
| $19.99 a month (recommended) | $9.99 a month | $14.99 a month |
| $99.99 a month | $49.99 a month | $74.99 a month |
| $199.99 a month | $99.99 a month | $149.99 a month |
Paying yearly costs 10 months’ price, so you get two months free. After cohort 1, cohort 2 gets 25% off and cohort 3 gets 10% off. Then it’s full price. Prices are from the Boost page on 16 November 2026.
Is it private?
Yes. We don’t keep your prompts and we don’t train on them. The serving code will be open source, so you can see how it works and move to your own hardware whenever you want. There’s no lock-in.
How does it fit with the Off Grid AI apps?
Off Grid AI runs open models on your own phone and computer, for free. Boost runs the same open models on our GPUs for when you need a bigger model, more speed or an API. You choose where each model runs.
How do I get access?
- Pick a plan on the Boost page, monthly or yearly.
- Pay your first month on Stripe’s checkout. That locks in the cohort price.
- Your API key arrives by email in January 2027, and your plan starts then.
FAQ
Is Boost live today?
Not yet. You can pre-order now to lock in the cohort price. Access and billing for your plan start in January 2027.
Which models are on Boost?
Open-weight models, including Qwen 3.8 and DeepSeek V4.1 Flash, across text, vision, speech, voice, embeddings, image and video. One key works for all of them.
Do I still need the Off Grid AI app?
No. Boost is an API. The Off Grid AI apps are free and run models on your own devices, with or without Boost.
Who makes Boost?
The team behind Off Grid AI, which has more than 250,000 downloads and more than 3,500 stars on GitHub.