OpenHuman with DeepSeek / Chinese Models — Cut API Costs by 90%

One of OpenHuman's best features is model flexibility. You can use any LLM provider, including affordable Chinese models like DeepSeek and Qwen (Tongyi Qianwen) that cost a fraction of OpenAI.

Why Use Chinese Models?

• Cost — DeepSeek V3 is about 90% cheaper than GPT-4o
• Performance — Chinese models are highly competitive, especially for logical reasoning
• Open source — many Chinese models are open-weight, allowing local deployment

Setting Up DeepSeek

DeepSeek offers an OpenAI-compatible API, making integration straightforward:

[models] [models.deepseek] provider = "openai" api_key = "sk-your-deepseek-api-key" base_url = "https://api.deepseek.com/v1" model = "deepseek-chat"

Using DeepSeek as the Default Model

In config.toml, set DeepSeek as your primary reasoning model:

[model_routing] reasoning_model = "deepseek" fast_model = "deepseek"

Setting Up Qwen (Tongyi Qianwen)

Alibaba's Qwen model family is another excellent choice:

[models.qwen] provider = "openai" api_key = "sk-your-qwen-api-key" base_url = "https://dashscope.aliyuncs.com/compatible-mode/v1" model = "qwen-plus"

Cost Comparison