Skip to content

Model Selection

Aigentic supports 25 AI models from 7 providers, including free cloud models and local models via Ollama. This guide covers the Models page, all available models, and how to choose the right one.

Open Models from the sidebar to access the model configuration page. It has two main sections:

Models are organized into a grid of provider cards — one per provider. Each card shows:

  • Provider name with a color-coded icon
  • Model count (e.g., “4 models available”)
  • “Free” badge on the NVIDIA NIM card
  • Model dropdown — Select a model from this provider to see its details
  • Model details panel — Shows the selected model’s tier badge, cost, “Recommended” badge (if applicable), and description
  • Test Connection button — Verifies that the provider is reachable with your current credentials. Shows “Connected” or “Connection Failed.”
  • Configure button — Navigates to the Credentials page to add or update your API key for this provider

Below the provider cards, a Fallback Chain card lets you configure what happens when the primary model is unavailable:

  • The chain shows numbered badges for each fallback model, connected by arrows
  • Click X on any badge to remove it from the chain
  • Use the Add model to chain dropdown to append a new fallback
  • Click Save Changes (appears when you have unsaved changes) to persist your configuration

These models are available immediately — no API key, no signup, no cost.

ModelProviderTierStrengths
DeepSeek V3.2NVIDIA NIMPowerful685B flagship. State-of-the-art open reasoning.
DeepSeek V3.1 TerminusNVIDIA NIMBalanced128K context, optimized for tool calling.
Kimi K2.5NVIDIA NIMPowerful1T multimodal model. Reasoning and vision.
Kimi K2 InstructNVIDIA NIMBalancedMoE architecture, fast and efficient.
GLM-5NVIDIA NIMPowerful744B MoE reasoning model. 205K context, tool calling.
GLM-4.7 FlashZ.AIFastFree, fast responses for simple tasks.
ModelTierCostStrengths
Claude Opus 4.6Powerful$5/1M tokensMost capable. 1M context, adaptive thinking.
Claude Opus 4.5Powerful$5/1M tokensPrevious-gen flagship. Strong reasoning and analysis.
Claude Sonnet 4Balanced$3/1M tokensBest balance of speed, quality, and cost.
Claude 3.5 HaikuFast$0.25/1M tokensFast and affordable for simple tasks.

API Key: console.anthropic.com

ModelTierCostStrengths
OpenAI o1Powerful$15/1M tokensAdvanced reasoning for complex problems.
OpenAI o3 MiniBalanced$1.1/1M tokensCompact reasoning model.
GPT-4oBalanced$2.5/1M tokensGeneral purpose, multimodal.
GPT-4o MiniFast$0.15/1M tokensFast and cost-effective.

API Key: platform.openai.com

ModelTierCostStrengths
Grok 4.1 Fast ReasoningBalanced$0.2/1M tokensFast reasoning with real-time knowledge.
Grok 3 MiniFast$0.3/1M tokensLightweight and fast.

API Key: console.x.ai

ModelTierCostStrengths
GLM-5Powerful$1/1M tokensFlagship reasoning and coding via Z.AI API.
GLM-4.7Balanced$0.6/1M tokensStrong all-rounder with 200K context.
GLM-4.7 FlashFastFreeFree, fast responses for simple tasks.

API Key: open.bigmodel.cn

ModelTierCostStrengths
MiniMax M2.5Powerful$0.15/1M tokensNear-SOTA coding and agents. 1/20th the cost of frontier.
MiniMax M2.5 HighspeedFast$0.3/1M tokens2x faster throughput. Optimized for real-time agents.

API Key: platform.minimaxi.com

Local models run on your machine via Ollama. Free, no API key, no data leaves your device. Requires the desktop app with Ollama installed.

ModelTierStrengths
Llama 3.3BalancedLatest Llama, runs locally. Free, no API key needed.
Llama 3BalancedRun locally, no API key needed.
MistralFastFast local model.
Code LlamaBalancedOptimized for coding tasks.
Phi-4FastCompact but capable.

Every model belongs to one of three tiers, shown as color-coded badges in the UI:

TierColorBest For
FastGreenQuick responses, simple tasks, high-volume workloads
BalancedBlueGeneral purpose, coding, analysis — best speed-to-quality ratio
PowerfulPurpleComplex reasoning, creative work, research, deep analysis

Recommended: DeepSeek V3.2 or Kimi K2.5 (NVIDIA NIM)

Start with free models to explore the platform. These are powerful models with no API key required.

Recommended: Claude Sonnet 4 or GPT-4o

Best balance of speed, capability, and cost. They handle a wide range of tasks well.

Recommended: Claude Opus 4.6 or GLM-5

When your agent needs deep analysis, synthesis, or nuanced output.

Recommended: Claude Sonnet 4 or MiniMax M2.5

Claude Sonnet 4 excels at code generation and review. MiniMax M2.5 offers near-SOTA coding performance at a fraction of the cost.

Recommended: MiniMax M2.5 ($0.15/1M) or GPT-4o Mini ($0.15/1M)

Best cost-per-token for agents processing high volumes of messages.

There are two ways to change which model an agent uses:

  1. On the Agents page, click the model name (with ✎ icon) on the agent card.
  2. Select a new model from the dropdown.
  3. Click Save.

The new model takes effect on the next message. Conversation history is preserved.

  1. Open Models from the sidebar.
  2. Select a model from any provider card dropdown.
  3. Click Save Changes.

This sets the default model for new agents and the primary model for the platform.

  • Start with free models — Use NVIDIA NIM or GLM-4.7 Flash while developing and refining your agents.
  • Configure fallbacks — Set up a fallback chain on the Models page so agents automatically switch to an alternative if the primary model is unavailable.
  • Monitor usage — The analytics dashboard tracks token consumption per agent.
  • Match model to task — Not every agent needs the most powerful model. A simple Q&A bot works well with a fast-tier model.
  • Use MiniMax M2.5 for cost-effective power — At $0.15/1M tokens, it offers near-frontier performance at 1/20th the price.