Models
Every model here is a GGUF file, checked before it is listed, and billed per token the model generates. Input tokens are not billed.
Models and prices
The app shows the same list, with the price in force and what is being served, before you send anything.
| Model | File size | Price |
|---|---|---|
| Gemma 4 12B Uncensored | 6.9 GB | shown in the app |
| LFM2 1.2B Extract | 697 MB | 1.00 USDC per 1M generated tokens |
| LFM2 1.2B RAG | 697 MB | 1.00 USDC per 1M generated tokens |
| LFM2 2.6B Transcript | 1.5 GB | 2.00 USDC per 1M generated tokens |
| LFM2.5 350M | 219 MB | 1.00 USDC per 1M generated tokens |
| Llama 3.2 3B Uncensored | 2.1 GB | 3.00 USDC per 1M generated tokens |
| Qwen2 Math 1.5B | 940 MB | 1.00 USDC per 1M generated tokens |
| Qwen2.5 1.5B Instruct | 1.0 GB | 1.00 USDC per 1M generated tokens |
| Qwen3.5 9B Uncensored | 5.2 GB | shown in the app |
Prices are set per model and can change. A request is answered by an operator who has the model loaded, so what is available moves as operators come and go. You can see both in the app before you send anything.
How models are added
- Upload
- A GGUF file, sent with a key. The file hash is the model id, so two uploads of the same file are the same model.
- Check
- A model is listed only after it has been checked. An unchecked model appears in no listing and serves no request.
- Price
- Set per model, per generated token, by the operator who publishes it. The app answers with the price in force before you send.
- Serving
- A worker loads a model onto its own hardware. The network owns no GPUs, so whether a model answers depends on an operator running it at that moment.
Serving these models
A worker earns 80% of what each job it serves is charged. Any model in the catalogue can be served by any machine that can hold it.