One unified API for all major Chinese LLMs. DeepSeek, Qwen3, Kimi K3, GLM-4 — 60%+ cheaper than official pricing.
If you're using api.deepseek.com, you need to migrate before the deadline.
Read the Migration Guide →
| Model | Input/M | Output/M | Official | Save |
|---|---|---|---|---|
| DeepSeek V3 | ¥1.0 | ¥3.0 | ¥2/¥8 | 62% |
| Qwen3 Max | ¥0.8 | ¥3.0 | ¥2/¥8 | 60% |
| Kimi K3 | ¥0.4 | ¥1.6 | ¥1/¥4 | 60% |
| GLM-4 | ¥0.4 | ¥0.4 | ¥1/¥1 | 60% |
| Qwen3 235B | ¥0.3 | ¥1.0 | ¥0.7/¥2.8 | 57% |
| Qwen3 32B | ¥0.15 | ¥0.6 | ¥0.5/¥2 | 67% |
| GLM-4-Air | ¥0.04 | ¥0.04 | ¥0.1/¥0.1 | 60% |
| DeepSeek R1 | ¥1.5 | ¥6.0 | ¥4/¥16 | 62% |
671B MoE, 128K context. HumanEval 89.2%. Best code generation among Chinese LLMs.
C-Eval 91.2%. 100+ languages. Best Chinese language understanding and generation.
1T MoE params, 128K context. Hybrid thinking mode. Unbeatable price-to-performance ratio.
262K context window. 235B MoE open-weight model. Perfect for long document analysis.
Just ¥0.04/M tokens. Perfect for simple classification, extraction, and formatting tasks.
GSM8K 95.2%. Specialized reasoning model for complex math, logic, and analysis tasks.
Complete API docs, integration guides, and code examples in Python, Node.js, and cURL.
Step-by-step guides for migrating from OpenAI, DeepSeek, DashScope, and other providers.
Open-source code examples, benchmarks, and community resources.
One API key. All Chinese LLMs. OpenAI-compatible. Start in 2 minutes.
Get Your API Key →