Flagship agent
Claude Fable 5
A high-capability model for difficult reasoning, complex code, and long tool workflows.
Best for
Complex agents, deep reasoning, long workflows
Back to API Platform6 Direct connection
PIXMIND API · AGENT INTEGRATIONS
One API key for the agents you already use
Connect PixMind GPT-5.6 models to coding agents, IDEs, and desktop clients through an OpenAI-compatible API. Choose a client and model here, then follow its focused setup guide.
OpenAI compatibleStreamingTool callingUnified model IDs
Shared Base URL
OpenAI Chat Completions
https://aihub-admin.aimix.pro/api-platform/v1
Anthropic
5Claude Fable 5 · Claude Opus 5 · Claude Sonnet 5 · Claude Opus 4.8 · Claude Opus 4.8 CC
DeepSeek
2DeepSeek V4 Pro · DeepSeek V4 Flash
智谱 AI
2GLM-5.3 · GLM-5.3-Flash
OpenAI
3GPT-5.6 Sol · GPT-5.6 Terra · GPT-5.6 Luna
Google
6Gemini 3.1 Pro Preview · Gemini 3.5 Flash · Gemini 3.6 Flash · Gemini 3.8 Flash · Gemini 3.7 Flash · Gemini 3.5 Flash-Lite
18 live models
SUPPORTED CLIENTS
Works with the agent tools you know
OpenAI Chat Completions clients connect directly. Claude Code uses PixMind's native Anthropic Messages endpoint, while Codex CLI connects through CC Switch local routing.
OPDirect connection
OpenCode
Add a custom OpenAI-compatible provider for terminal and desktop coding agents.
OpenAI Chat CompletionsRead setup guide
CUDirect connection
Cursor
Use the OpenAI API key and Base URL override in Models settings.
OpenAI Chat CompletionsRead setup guide
TRDirect connection
Trae
Add an OpenAI-compatible custom model in model management.
OpenAI Chat CompletionsRead setup guide
CHDirect connection
Cherry Studio
Add an OpenAI-compatible provider and reuse models in chats and agents.
OpenAI Chat CompletionsRead setup guide
CLDirect connection
Cline / Roo Code
Choose OpenAI Compatible in the VS Code extension and enter a model ID.
OpenAI CompatibleRead setup guide
COVia CC Switch
Codex CLI
Uses CC Switch local routing to convert Responses into PixMind Chat Completions.
Responses → CC Switch → Chat CompletionsRead setup guide
CLDirect connection
Claude Code
Connects directly through the Anthropic Messages-compatible endpoint with model discovery, token counting, and tools.
Anthropic MessagesRead setup guide
MODEL ROUTING
Match the model to the workload
PixMind offers language models from OpenAI, Anthropic, and Google through one API and authentication method. Choose by reasoning, coding, multimodal input, speed, and cost.
Anthropic
5 live modelsView model details
Advanced reasoning
Claude Opus 5
Built for complex analysis, architecture decisions, and sustained agent reasoning.
Best for
Architecture, complex planning, difficult debugging
Balanced coding
Claude Sonnet 5
Balances coding quality, tool use, and response efficiency.
Best for
Daily coding, tool calls, code review
Complex engineering
Claude Opus 4.8
Suited to large engineering analysis, cross-file refactors, and difficult debugging.
Best for
Large codebases, cross-file refactors, hard problems
Coding agent
Claude Opus 4.8 CC
A Claude route focused on coding agents and tool-driven tasks.
Best for
Coding agents, terminal tools, development automation
DeepSeek
2 live modelsGeneral model
DeepSeek V4 Pro
Best for
General model
DeepSeek V4 Flash
Best for
智谱 AI
2 live modelsGeneral model
GLM-5.3
Best for
General model
GLM-5.3-Flash
Best for
OpenAI
3 live modelsFlagship reasoning
GPT-5.6 Sol
The flagship tier for complex engineering, long-context analysis, and difficult planning.
Best for
Large codebases, architecture, complex debugging
Recommended
Balanced default
GPT-5.6 Terra
A balanced mix of quality, speed, and cost for everyday agent work.
Best for
Daily development, tool calls, mixed workloads
Fast and light
GPT-5.6 Luna
Low-latency responses for fast edits, search, and frequent lightweight tasks.
Best for
Quick answers, small edits, batch tasks
Multimodal reasoning
Gemini 3.1 Pro Preview
For multimodal tasks that combine image and video understanding with complex reasoning.
Best for
Complex multimodal analysis, image and video understanding
Balanced multimodal
Gemini 3.5 Flash
Balances multimodal input, tool calls, and responsive Flash performance.
Best for
Multimodal agents, visual Q&A, tool calls
Efficient multimodal
Gemini 3.6 Flash
Processes text, images, video, and tool calls with higher efficiency.
Best for
Frequent multimodal work, video understanding, batch processing
Long-horizon multimodal agents
Gemini 3.8 Flash
Built for long-horizon software engineering, autonomous agents, and complex multimodal work.
Best for
Long coding tasks, multimodal agents, complex tool workflows
Vision and video
Gemini 3.7 Flash
Designed for image and video understanding with fast multimodal responses.
Best for
Visual analysis, video Q&A, fast content understanding
Light multimodal
Gemini 3.5 Flash-Lite
A lightweight option for frequent multimodal processing and extraction.
Best for
Classification, extraction, summaries, light vision tasks
GET STARTED
Start in three steps
- 01
Create an API key
Create a key in PixMind API Platform and store it only in a secure credential store or environment variable.
- 02
Open the client guide
Set the Base URL, API key, and model ID. OpenCode stores the key separately through auth login.
- 03
Choose a model and verify
Start with GPT-5.6 Terra and a minimal prompt, then adjust the model and reasoning level.
Codex CLI can use PixMind through CC Switch
Codex CLI speaks Responses while PixMind currently exposes Chat Completions. CC Switch 3.17.0+ local routing converts requests, streaming responses, and tool calls in both directions, so Codex can use PixMind through CC Switch rather than a native Responses connection.
Choose an agent and start building
Every directly supported client shares one Base URL, API key, and set of model IDs.