Supported LLM Providers

DuRT adopts a strict BYOK (Bring Your Own Key) architecture with zero platform markups. You directly connect to the model provider of your choice, paying only for the tokens you consume.

---

Supported Cloud LLM Providers

Provider	Supported Models	Base URL Configuration	Key Features
OpenAI	GPT-4o, GPT-4o mini, o1	https://api.openai.com/v1	Industry benchmark, official Realtime API support
Anthropic Claude	Claude 3.5 Sonnet, Claude 3.5 Haiku	https://api.anthropic.com/v1	Superior literary nuance and translation quality
Google Gemini	Gemini 1.5 Pro, Gemini 1.5 Flash	https://generativelanguage.googleapis.com	Million-token context window for full-length books
DeepSeek	DeepSeek-V3, DeepSeek-R1	https://api.deepseek.com/v1	Extreme cost efficiency, exceptional reasoning
Groq	Llama 3.3 70B, Mixtral	https://api.groq.com/openai/v1	Ultra-fast LPU inference (500+ tokens/sec)
xAI Grok	Grok-2	https://api.x.ai/v1	Real-time information and creative writing
Mistral AI	Mistral Large, Codestral	https://api.mistral.ai/v1	European sovereign AI, multilingual proficiency
OpenRouter	200+ models aggregated	https://openrouter.ai/api/v1	Single key access to open & closed models
Zhipu AI (GLM)	GLM-4 Plus, GLM-4 Air	https://open.bigmodel.cn/api/paas/v4	Native Chinese context and document reasoning
Aliyun Qwen	Qwen-2.5-72B, Qwen-Max	https://dashscope.aliyuncs.com/compatible-mode/v1	Comprehensive multilingual accuracy
Moonshot (Kimi)	Moonshot-v1	https://api.moonshot.cn/v1	Long-context dialogue and document summary
SiliconFlow	DeepSeek, Qwen, Flux	https://api.siliconflow.cn/v1	High-concurrency domestic API gateway
Tencent Hunyuan	Hunyuan-Turbo	Tencent Cloud Endpoint	Enterprise Chinese linguistic nuances
Baidu Ernie	Ernie 4.0	Baidu Cloud Endpoint	Knowledge grounding and search augmentation

---

Local & Private Inference Engines

1. Ollama (Local First)

• Zero Cloud Leakage: Run models completely offline on Apple Silicon Macs.
• Default Endpoint: http://localhost:11434/v1
• Recommended Models: llama3.2, qwen2.5:7b, deepseek-r1:8b.

2. Custom OpenAI-Compatible Endpoints

• Connect any self-hosted inference server (vLLM, LocalAI, LM Studio, Ollama, FastChat).
• Simply provide the API Base URL, API Key (or dummy string), and custom Model Name.