Nexus
The right mind for every question.
Nexus is the smart router that selects the optimal AI model for every request. It classifies query type, estimates complexity, and routes to the best model based on capability, cost, and latency.
With 28+ supported models from OpenAI, Anthropic, Google, Meta, and others, choosing the right one matters. A simple question doesn't need a 405B parameter model. A complex coding task shouldn't go to a lightweight model. Nexus eliminates this decision entirely.
The routing system considers multiple dimensions: task category, complexity level, cost constraints from your plan, latency requirements, and provider availability. The result is optimal model selection without manual intervention.
Routing Intelligence
Query Classification
Automatic detection of task type: code generation, mathematical reasoning, creative writing, data analysis, general conversation, and more.
Complexity Estimation
Three-tier complexity scoring determines whether a simple, medium, or flagship model is needed for the task.
Cost-Aware Routing
Balance quality and cost. Simple questions route to efficient models, saving budget for complex tasks that need premium capability.
Latency Optimization
Real-time tasks route to faster models. Batch processing can use slower, more thorough models for better results.
Provider Preference
Set preferred providers and Nexus respects your choice while still optimizing within that constraint.
Capability Profiles
Each model has a detailed capability profile: strengths, weaknesses, token limits, and performance benchmarks by task type.
Provider Failover
If the primary provider is down, Nexus automatically switches to a fallback (e.g., OpenRouter to Groq). Conversations continue seamlessly.
Usage Learning
Nexus learns from your usage patterns. Models that consistently produce better results for your tasks get prioritized over time.
Multi-Model Support
28+ models: GPT-5, Claude Opus, Gemini 3.0, Llama 4, DeepSeek V4, Qwen 3, Mistral Large, Kimi K2.5, and more via OpenRouter.
How Nexus routes a query
Query arrives
Your message enters the system. Nexus analyzes the content to determine task type and required capabilities.
Classification & scoring
The query is classified by category and complexity. Each available model receives a score based on capability match, cost, and latency.
Model selected
The highest-scoring model is selected, respecting your plan limits and provider preferences. Fallback options are queued.
Result feedback
After execution, the quality of the model's response feeds back into the scoring system, improving future routing decisions.