Model catalog
Open-source model catalog
Filter open models by size, memory need and license — we deploy any of them privately, on your hardware.
| Model | Params | Min VRAM | License | Best for |
|---|---|---|---|---|
| Phi-3.5 mini | 3.8B | 4 GB | MIT | Edge & on-device |
| Mistral 7B | 7B | 6 GB | Apache 2.0 | Fast assistants |
| Llama 3.1 8B | 8B | 8 GB | Llama 3.1 | Copilots & chat |
| Qwen 2.5 7B | 7B | 8 GB | Apache 2.0 | Multilingual chat |
| Gemma 2 9B | 9B | 10 GB | Gemma | Lightweight tasks |
| Gemma 2 27B | 27B | 20 GB | Gemma | Balanced quality |
| Qwen 2.5 32B | 32B | 24 GB | Apache 2.0 | Reasoning & analysis |
| Mixtral 8x7B | 47B MoE | 24 GB | Apache 2.0 | Throughput & RAG |
| Llama 3.3 70B | 70B | 40 GB | Llama 3.3 | High-quality reasoning |
| Qwen 2.5 72B | 72B | 40 GB | Qwen | Complex tasks |
| Mixtral 8x22B | 141B MoE | 80 GB | Apache 2.0 | Enterprise workloads |
| Llama 3.1 405B | 405B | Multi-GPU | Llama 3.1 | Maximum quality |
| DeepSeek R1 | 671B MoE | Multi-GPU | MIT | Frontier reasoning |
| Falcon 180B | 180B | Multi-GPU | Falcon | Large-scale serving |
No models match your filter.
Indicative minimum VRAM for quantized inference; we size the exact hardware per use case.