Models
The best open models, served fastest
Run leading open-source and frontier models through one OpenAI-compatible API — no infrastructure to manage, no rate-limit surprises.
| Model | Provider | Context | Speed | Type |
|---|---|---|---|---|
Llama 3.1 405BFlagship | Meta | 128K | 132 tok/s | Chat |
Llama 3.3 70BFast | Meta | 128K | 460 tok/s | Chat |
DeepSeek-V3New | DeepSeek | 64K | 250 tok/s | Reasoning |
DeepSeek-R1New | DeepSeek | 64K | 198 tok/s | Reasoning |
Qwen 2.5 72B | Alibaba | 128K | 410 tok/s | Chat |
gpt-oss 120BFastest | Open | 32K | 520 tok/s | Chat |
E5 Mistral Embed | Open | 8K | — | Embedding |
Speeds measured on Zerafort SN50 RDUs in production. New models added every week.