AI models available through the gateway
Available Models
Vindex Ai provides access to high-performance AI models through Groq.
Model Overview
The following models are available. The model is selected when creating or editing a site.
| Model | Context | Max Output | Speed | Best For |
|---|---|---|---|---|
| Llama 3.3 70B | 128K | 8K | ~180 TPS | General purpose, reasoning |
| Llama 3.1 70B | 128K | 8K | ~150 TPS | Complex tasks, code generation |
| Llama 3.2 90B | 128K | 8K | ~100 TPS | Large context tasks |
| Mixtral 8x7B | 32K | 32K | ~200 TPS | Fast responses, coding |
| Gemma 2 9B | 8K | 4K | ~250 TPS | Lightweight, fast tasks |
Selecting a Model
For General Use
Recommended: Llama 3.3 70B
Best for: Customer support, general conversation, Q&A
For Code Generation
Recommended: Llama 3.1 70B or Mixtral 8x7B
Best for: Programming help, code review, technical questions
AI Architect (Scribe)
Writing the perfect system prompt is the most important part of building an elite AI channel. To help you with this, Vindex provides Scribe, our built-in AI Architect Agent.
How Scribe Helps You
- The Interviewer: Scribe won't just guess your needs. It will interview you about your AI's goal, tone, and "Hard Rules" (things it should never say).
- Template Library: Scribe has access to high-performance, industry-tested templates for Support, Sales, Real Estate, and more.
- Quality Analysis: Every prompt Scribe generates is automatically analyzed for clarity, structure, and security.
- Instant Deployment: Once you are happy with a prompt, Scribe can apply it to your channel with a single click.
To work with Scribe, simply ask the Nexus Assistant (Tenant Assistant) or Vin (Admin Assistant) to "Connect me with Scribe" or "Help me write a new prompt."