Setup guide

Provider setup guide

Browse setup steps and troubleshooting by tool.

Provider setup guide/Models and vendors: First know whose model you are choosing

Models and vendors: First know whose model you are choosing

First differentiate between vendors, model series, and model IDs, know what each type of model is suitable for, and then choose the AI provider and tool tutorials.

First, distinguish the three names

NamePlain explanationHow to view when configuring
VendorCompanies or platforms that develop or provide model services, such as OpenAI, Anthropic, Google, DeepSeek, Alibaba Cloud, ByteDance and Volcano Engine, etc.Manufacturers decide on model capabilities, interface protocols, prices, and limitations; the AI provider simply connects these capabilities to its own backend.
Model seriesA set of model brand or family names, such as GPT, Claude, Gemini, DeepSeek, Qwen, GLM, Doubao.The series name only helps you determine the origin and general capability; it cannot be directly equated with the model ID that needs to be filled in the configuration.
Model ID / Model NameThe exact name that needs to be filled in the API request, such as the string of characters in the provider's backend model list.Copy exactly from the backend, do not change the case, hyphens, periods, or version numbers yourself; if obtaining the model list fails, manually add the model IDs provided by the backend.
Display NameThe Chinese name or alias displayed in the tool or service backend for easier reading.The display name may not be directly callable; when configuring, prioritize fields like 'Model ID', 'Model', or 'Model Name'.

Common Foreign Model Vendors

VendorCommon Models or SeriesCommon Uses
OpenAIGPT, o series, GPT Image, Sora, etc.Text conversations, code, reasoning, image generation, video generation
AnthropicClaude SeriesLong text reading, code, writing, analysis
GoogleGemini SeriesMultimodal understanding, long context, code, image understanding
xAIGrok SeriesText conversations, reasoning, real-time information scenarios
MetaLlama SeriesOpen-source weights, self-deployment, third-party hosting
Mistral AIMistral, Codestral, etc.Text conversations, code, lightweight model scenarios
CohereCommand, Embed, Rerank, etc.Enterprise knowledge base, retrieval augmentation, vectors and reranking
Stability AIStable Diffusion, Stable Image, etc.Image generation, image editing
RunwayGen series video modelsVideo generation, video editing

Common Domestic Model Vendors

VendorCommon Models or SeriesCommon Uses
DeepSeekDeepSeek Chat, DeepSeek Reasoner, etc.Text conversations, code, reasoning
Alibaba Cloud / QwenQwen, Qwen-VL, Qwen-Coder, etc.Text conversation, image understanding, code, open-source models
Moonshot AIKimi, Moonshot SeriesLong text reading, writing, and data analysis
Zhipu AIGLM, GLM-4, etc.Text dialogue, reasoning, multimodal, intelligent agent
ByteDance / Volcano Engine ArkDoubao seriesText dialogue, multimodal, vectors, enterprise applications
Baidu AI Cloud / QianfanERNIE / Wenxin SeriesText conversation, multimodal, search, and enterprise scenarios
Tencent CloudHunyuan seriesText chat, multi-modal, enterprise access
MiniMaxMiniMax, abab, voice-related modelsText conversation, long text, voice, and multimodal
iFlytekSpark SeriesText conversations, educational office, voice-related scenarios
01. AIYi SeriesOpen-source models, text dialogue, third-party hosting

Common Model Types

TypeWhat does it doPay attention when configuring
Chat / Text ModelAnswer questions, write copy, translate, summarize, write code.Most conversational tools prioritize configuring this type of model, making it most suitable for the first step of testing.
Reasoning modelsIt is more suitable for complex analysis, mathematics, code reasoning, and multi-step tasks.The price and response time may be higher, so simple chats don't necessarily require priority.
Code modelIt leans more towards reading code, writing code, fixing bugs, and explaining projects.In code tools, also check whether tool calls, file read/write, and context management are supported.
Visual understanding modelsView images, recognize screenshots, understand tables or interfaces.The tool itself should support image uploads, and providers should also allow visual input.
Image generation modelGenerate or edit images based on prompts.Usually not /chat/completions For this kind of ordinary chat interface, the configuration page should be based on the image API.
Video generation modelGenerate short videos, image-to-video videos, or video editing.The common process is submitting the task, waiting for completion, then retrieving the result, rather than instant chat replies.
Embedding vector modelConvert text into vectors for knowledge base retrieval and similarity matching.It cannot be used for chatting; it is usually used together with knowledge bases, RAG, and search systems.
Rerank rearranged the modelReorder search results so that more relevant content is placed at the top.It is commonly used for optimizing knowledge base performance and is not a model for generating answers.
Voice modelsText-to-speech, voice-to-text, real-time voice conversation.Fields differ from regular text models; you need to refer to the corresponding TTS / STT / Realtime documentation.

How do beginners choose?

What are you going to do?What should you prioritize?Reason
General chat, translation, summarizationStable, low-cost chat modelsGetting the Key, API address, and model name working is more important than initially pursuing the most powerful model.
Writing code, modifying projectsModels with strong coding or reasoning abilitiesCoding tasks rely more on context, tool calls, and accuracy; cheaper models are prone to repeated revisions.
Reading long documents, long conversationsText models with longer contextLonger context allows for more information, but the cost may also be higher.
Viewing screenshots, analyzing imagesMultimodal models that support visual inputWhen only a regular text model is configured, uploading images may be unavailable or not understood.
Generating imagesImage generation modelImage generation is not the same as a chat model; usually, you need to consider image tools and image APIs separately.
Generating videosVideo generation models or video platformsVideo models consume a lot and have long wait times; first, check the billing and task process carefully.
Creating a knowledge baseChat model + Embedding + optional RerankA knowledge base does not rely solely on a chat model; retrieval and re-ranking significantly affect answer quality.
If you are configuring tools in an AI provider, the most stable order is: first choose an inexpensive text model and get it working, then test the code, vision, image, or video model you actually need. Do not fill a model name from one provider into another provider's protocol just because the model series names are similar.