Model Library
Ready-to-go open models, pre-configured for one RTX 5090. Click Deploy, pick a GPU, and get an API or web UI in minutes — or bring your own model with a template.
Reasoning
Qwen3 8B
Dense 8B model with switchable thinking / non-thinking modes, strong multilingual chat and tool use.
Qwen3 14B (FP8)
Larger Qwen3 in official FP8 weights — noticeably stronger reasoning while fitting comfortably on one GPU.
Qwen3 32B (AWQ)
The flagship dense Qwen3, 4-bit AWQ quantised to run on a single RTX 5090.
Qwen2.5 72B Instruct (AWQ)
A 70B-class model split across both GPUs of a 2× RTX 5090 machine (64 GB VRAM) with tensor parallelism.
DeepSeek R1 Distill Qwen 7B
Small reasoning model distilled from DeepSeek-R1 — shows its chain of thought, strong at maths.
gpt-oss 20B
OpenAI's open-weight reasoning model with adjustable reasoning effort, in a ready-to-use chat UI.
DeepSeek R1 Distill 32B (Ollama)
The 32B R1 distill in 4-bit via Ollama — near-frontier reasoning on a single GPU.
Phi-4 14B (Ollama)
Microsoft's compact model trained on high-quality synthetic data — strong maths and reasoning for 14B.