Model Comparison
Compare capabilities, pricing, and use cases of each model to find the best fit for your needs.
Chat Models
| Model | Provider | Context Length | Input Price (per 1M tokens) | Output Price (per 1M tokens) | Key Strengths |
|---|---|---|---|---|---|
| deepseek-v4-flash | DeepSeek | 1M | ¥1.0 | ¥2.0 | Fastest response, lowest cost |
| deepseek-v4-pro | DeepSeek | 1M | ¥12.0 | ¥24.0 | Complex reasoning, highest quality |
| qwen3.7-max | Qwen | 1M | ¥2.0 | ¥6.0 | General conversation, strong Chinese NLU |
| glm-5.7 | Zhipu AI | 200K | ¥2.0 | ¥5.0 | General conversation, knowledge QA |
| kimi-k2.6 | Moonshot AI | 256K | ¥4.0 | ¥12.0 | Ultra-long context, document analysis |
| minimax-m3 | MiniMax | 1M | ¥2.0 | ¥8.0 | General conversation, multimodal |
Choosing a Chat Model
| Use Case | Recommended Model | Reason |
|---|---|---|
| Simple Q&A, chatbots | deepseek-v4-flash | Fastest and cheapest |
| General conversation, content | qwen3.7-max | Best value, strong Chinese understanding |
| Complex reasoning, code gen | deepseek-v4-pro | Strongest reasoning, highest output quality |
| Long document analysis | kimi-k2.6 | 256K context, excels at document understanding |
| Knowledge QA, education | glm-5.7 | Broad knowledge coverage, stable responses |
| Multimodal conversation | minimax-m3 | Supports multiple input modalities |
Image Generation Models
| Model | Provider | Price (per image) | Max Resolution | Key Strengths |
|---|---|---|---|---|
| Kolors | Kuaishou | ¥0.01 | 1024×1024 | Ultra-low cost, ideal for batch gen |
| gpt-image-2 | OpenAI | ¥0.06 | 1536×1024 | High quality, versatile, text rendering |
| nano-banana-2 | — | ¥0.12 | 1024×1024 | Photorealistic style |
| doubao-seedream-4.0 | ByteDance | ¥0.20 | 1024×1024 | Chinese art styles |
| doubao-seedream-5.0-lite | ByteDance | ¥0.22 | 1024×1024 | Upgraded Doubao, faster |
| doubao-seedream-4.5 | ByteDance | ¥0.25 | 1024×1024 | High-quality Doubao, richer details |
Choosing an Image Generation Model
| Use Case | Recommended Model | Reason |
|---|---|---|
| Batch generation, testing | Kolors | Lowest cost, great for rapid iteration |
| General high-quality images | gpt-image-2 | Best overall quality, strong text render |
| Photorealistic people/product | nano-banana-2 | Excellent photorealistic results |
| Chinese art / illustrations | doubao-seedream series | Better understanding of cultural elements |
Video Generation Models
| Model | Provider | Supported Modes | Max Resolution | Max Duration | Key Strengths |
|---|---|---|---|---|---|
| veo-3 | Text-to-video, image-to-video | 1920×1080 | 8 sec | Top-tier quality, cinematic results | |
| wanx2.1-t2v-turbo | Alibaba | Text-to-video, image-to-video | 1920×1080 | 10 sec | Diverse styles, great value |
| doubao-seedance-2-0 | ByteDance | Text-to-video, image-to-video, audio-driven | 1920×1080 | 10 sec | Digital avatars, audio-driven lip sync |
Choosing a Video Generation Model
| Use Case | Recommended Model | Reason |
|---|---|---|
| Highest quality, brand content | veo-3 | Best visual quality and motion dynamics |
| Short-form video, ad creatives | wanx2.1-t2v-turbo | Diverse styles, up to 10s, great value |
| Digital avatars, news anchor | doubao-seedance-2-0 | Audio-driven lip sync, built for avatars |
| Image-to-video, photo animation | Any of the above | All support image-to-video; choose by quality need |
Web Search
| Model | Type | Price (per request) | Description |
|---|---|---|---|
| web-search | Search-augmented | ¥0.02 | Real-time internet retrieval, Chat-compatible |
Web search can be used independently or combined with chat models (via function calling or tool use) to give AI access to the latest real-time information.
Overall Comparison
By Price (Low to High)
| Category | Most Affordable | Mid-Range | Premium |
|---|---|---|---|
| Chat | deepseek-v4-flash | qwen3.7-max | deepseek-v4-pro |
| Image | Kolors | gpt-image-2 | doubao-seedream series |
| Video | wanx2.1-t2v-turbo | veo-3 | doubao-seedance-2-0 |
By Use Case
| Scenario | Recommended Stack |
|---|---|
| Solo developer / testing | deepseek-v4-flash + Kolors + web-search |
| Content creation team | qwen3.7-max + gpt-image-2 + wanx2.1-t2v-turbo |
| Enterprise applications | deepseek-v4-pro + gpt-image-2 + veo-3 + web-search |
| Short-form video / avatars | qwen3.7-max + doubao-seedream + doubao-seedance-2-0 |
How to Choose the Right Model
- Prioritize your requirements: Determine whether quality, speed, or cost matters most
- Start small: Validate your workflow with a lower-cost model, then switch to a higher-quality model once confirmed
- Use free credits: New users receive ¥5 in free credits — enough to test all models
- Stay updated: Models are continuously upgraded. Check the Model List regularly for the latest information
- Mix and match: Use different models for different tasks to achieve the best balance of cost and quality
Prices are subject to change. Refer to the Console for real-time pricing.