Skip to main content

Models & Pricing

Select the right level of intelligence for your agents from a diverse range of state-of-the-art AI models.

This guide outlines the available models, their credit costs, and how consumption is calculated in SVAHNAR.


💎 Available Models​

The following models are available for immediate use in your agent configurations.

ProviderModel Name (Config ID)Credits Per Run
OpenAIgpt-4o-mini1
gpt-4o15
GPT-5-nano1
GPT-5-mini3
GPT-515
LLaMALlama 3.1 8B instruct1
Llama 3.3 70B instruct2
DeepseekDeepseek - Qwen-2.5-32B2
Deepseek - LLama70B3
QwenQwen QwQ 32B (Preview)1
Qwen 2.5 32B Instruct 128k3
Qwen 2.5 Coder 32B Instruct 128k3
Enterprise

Need custom solutions? Enterprise customers can contact sales for access to fine-tuned models or custom deployments.


📚 Practical Recipes​

Setting the Model in Config​

To use a specific model, copy the Model Name from the table above and paste it into the model parameter of your YAML configuration.

Use Case: Configuring an agent to use a cost-effective model for simple tasks.

create_agent_network:
agent-1:
agent_name: simple_task_agent
LLM_config:
params:
# Paste the Model Name here
model: gpt-4o-mini
tools: []
agent_function:
- Reply to the user's greeting.

💡 Core Concepts​

Before selecting a model, it is important to understand how SVAHNAR calculates usage.

1. What is a "Run"?​

In a multi-agent network, pricing is calculated based on Runs. A Run occurs whenever an agent is triggered to perform a task or participate in a conversation.

2. How are credits consumed?​

Credits are deducted strictly on a pay-per-run basis.

  • If an agent uses gpt-4o-mini, 1 credit is deducted for that turn.
  • If an agent uses GPT-5, 15 credits are deducted for that turn.
info

In a complex workflow where multiple agents interact, the total cost is the sum of the credits required for each agent's individual participation.