*TLDR [NanoGPT](https://nano-gpt.com/) offers private access to every AI model you can think of. No subscription, pay only for what you use, text, image, video. No account needed. Reply for an invite, or deposit as little as $0.10 to try it out..*

# Why is this relevant for me?

If you want to use the top AI models (like o3-pro, Grok and Veo 3) but don’t want a $200 subscription, we’re the place to go.

If you want to use the top AI models but don’t like the idea of linking your name, personal details and credit card to the prompts you do, we’re the place to go.

If you want to use uncensored models, we’re the place to go.

If you care about crypto being actually used, we’re probably a place you want to check out and give a try.

## Improvements

– **Account deletion is now a lot easier**. This used to be manual which frankly, not great. You can now easily do it from Settings and optionally provide a reason for deletion, but there’s no need to do so.
– **Open source model mode**: toggle open source mode in [Settings](https://nano-gpt.com/settings) and all closed source text models will be hidden from the options. For all of the open source models that we offer we do NOT run them directly through the provider, because many of the best open source models are Chinese and frankly we understand that distrust. We use no-log providers and are largely moving to TEE.
– **TEE models**: we have a [blog up on this here](https://nano-gpt.com/blog/trusted-execution-environments-and-confidential-ai-on-nanogpt). They are the most private way to use open-source models online. We have a TEE category where the models are verifiably run through Trusted Execution Environments. For the newly added Qwen, Kimi, and GLM (see below) we also run them through a provider that claims TEE, but we have not been able to fully verify that yet so they’re not shown in the TEE category yet. Deepseek V3 for example can be run through TEE.
– **Price decreases**: for video and image models we should now be pretty much the cheapest or close to the cheapest provider for all of them. If you’re running text models via API we match the price of the direct providers or beat it, so let’s say ChatGPT costs $5 per mln tokens directly through OpenAI, we now charge $5 per mln tokens via API ourselves as well. For many of the open-source models we’re cheaper than any publicly available option. For anyone that wants to build AI into their application/website – deposit anonymously, don’t create an account, yet get the same or lower prices than at providers that want your credit card.

## New models

**Text models:**

– Grok 4.
– Vercel v0 models. Very useful for frontend development, especially next.js/Vercel applications. We’re the only ones that offer this as far as we know.
– [Study mode](https://x.com/NanoGPTcom/status/1951290272064627038)! ChatGPT with a custom instruction designed to help you learn through guided discovery rather than providing direct answers.
– Qwen 3 Coder, Qwen 3 235b a22b instruct + thinking.
– Kimi K2 (also via Groq for super-speedy inference).
– GLM 4.5.
– Step 3

**Video models**:

– Veo 3 image to video
– Veo 3 fast (cheaper), text to video and image to video
– Wan 2.2 (full, 5b and Turbo). Cheapest version here costs just $0.05 per video and has surprisingly good quality.
– PixVerse 4.5
– Vidu Q1
– Face swap video
– Longstories

The prices on our video models have also been much decreased. Using them directly via the website or via API we should now be the lowest or one of the lowest priced providers for all of them. We now also do automatic refunds for failed generations.

**Image models**

I’d love to list them all, but it’s genuinely too many. We added about 30, the biggest ones are Flux Krea, Hidream Edit E1.1, and Higgsfield Soul. Also face swap, which is quite fun to do.

Prices for our image models have also all been dropped.

That’s all – happy to answer questions, and would mostly be happy if this inspires some other merchants to also start accepting crypto.