Cherry Studio is the power-user's open-source AI desktop client: multi-provider, multi-assistant, knowledge bases, MCP support, and mini-apps — all against any OpenAI-compatible endpoint. Pointing it at NovAI gives one interface for the entire Chinese model zoo: chat, vision, image generation, and translation assistants side by side.
NovAI, choose type OpenAI.https://aiapi-pro.com/v1glm-5.3, kimi-k2.7-code, qwen3.8-max, glm-4.1v-thinking-flash, cogview-3-flash. (Cherry Studio can also pull the live list from /v1/models.)| Assistant | Model | Why |
|---|---|---|
| Daily chat | glm-4.7-flash | Free forever, fast |
| Deep work / long docs | glm-5.3 | 1M context |
| Coding | kimi-k2.7-code | Agent-tuned, $0.67/1M in |
| Image analysis / OCR | glm-4.1v-thinking-flash | Free visual reasoning |
| Image generation | cogview-3-flash / doubao-seedream-5.0 | Free tier or 2K quality at $0.027 |
| Translation | qwen3.5-flash | $0.048/1M in, high volume |
Cherry Studio's knowledge-base feature embeds documents against your chosen provider — use cheap qwen3.5-flash or deepseek-v4-flash ($0.08/$0.17) for embedding-heavy ingestion, then answer with glm-5.3. MCP servers configured in Cherry Studio work unchanged since tool-calling flows through the same OpenAI-compatible API.
| Model | Input $/1M | Output $/1M |
|---|---|---|
| kimi-k2.7-code (agent-tuned) | $0.67 | $3.40 |
| glm-5.3 (1M context flagship) | $1.25 | $4.00 |
| deepseek-v4-pro | $0.57 | $1.15 |
| doubao-seed-2.0-code | $0.448 | $2.24 |
| glm-4.7-flash (free forever) | $0 | $0 |
For long agent loops, kimi-k2.7-code and deepseek-v4-pro give the best cost-per-task; for repo-wide refactors that need the full codebase in context, glm-5.3's 1M window is unmatched. Use free glm-4.7-flash for planning and lightweight edits.
One desktop client, 40+ Chinese models — $2 free credit, four free forever.
In Settings > Model Provider, add a provider of type OpenAI with API URL https://aiapi-pro.com/v1 and your NovAI API key, then add model ids manually or fetch them from the /v1/models endpoint.
Yes. Add cogview-3-flash (free forever) or doubao-seedream-5.0 ($0.027/image) and use Cherry Studio's drawing assistants; note Chinese image models burn a visible AI-generated label into output.
Use a cheap model like qwen3.5-flash for ingestion and retrieval-heavy chats, and glm-5.3 with its 1M context for answering over large document sets.