Models
Three ways to get a model on endlesstokens: popular open-source models we host and add as demand grows, frontier models offered through their labs by agreement, and your own licensed models served for you.
Open-source models
Every model here runs on GPUs we host and scale ourselves. We benchmark each one and derive its unit measures before it goes on offer, then keep launching the next popular model as demand shows.
Qwen3.8-27B
- Open weights, hosted by us. Deployed and scaled on GPUs we operate, never proxied to another provider.
- Context: served at 128K tokens.
- Capabilities: vision (up to 5 images per request), reasoning with effort capped at medium, and tool calling.
The licence and the full specification are on the model card.
Want another open-source model?
Tell us which one. Every request counts toward our demand assessment, and the popular ones we host, benchmark and add to the catalogue. No dates are promised.
Frontier models
Closed, proprietary frontier models belong to the labs that build them. We do not proxy or resell them. When we agree terms with a lab, we will offer its models through that lab, on its terms and with its approval.
Not available today. Tell us which frontier model you would want and it informs which agreements we pursue.
Custom and proprietary models
Need a specific model that is not in the catalogue? If you hold the licence to run it, including closed-weight and proprietary models, we can serve it for you on the same platform. You confirm the licence, we host and scale it.