Your model choice strongly affects reply quality, latency, and cost. In Chatbot → AI Settings, the provider you pick determines which chat models appear in the list (supported services include OpenAI, Google Gemini, Claude (Anthropic), DeepSeek, Grok (xAI), and OpenRouter).
Key Point: Each provider offers different model families with different capability and price trade-offs. Use the tables below for orientation, then validate results against your real traffic and budget.
Understanding Model Types
Limb Chatbot uses different model types for different jobs. Knowing the distinction helps you choose settings for chat versus knowledge (embeddings):
Chat Models
Used for conversational AI—answering questions, generating replies, and similar text tasks. These are the models visitors interact with in the chat.
Embedding Models
Used for knowledge indexing—turning text into embedding vectors so the system can run semantic search. They power retrieval from your knowledge base, not the visible chat personality.
Chat Models Comparison
Chat models power visitor-facing conversations. The table below compares common options:
Model
Provider
Intelligence
Speed
Cost
Best For
gpt-4.1
OpenAI
Highest
Medium
$$$
Complex reasoning, technical support
gemini-2.5-pro
Gemini
Highest
Medium
$$$
High-quality responses, detailed content
gpt-5-mini
OpenAI
High
Fast
$$
Balanced – most general chatbots
gpt-4.1-mini
OpenAI
High
Fast
$$
General purpose, good quality
gemini-2.5-flash
Gemini
High
Very Fast
$$
Fast responses, customer support
gpt-4o
OpenAI
High
Fast
$$
Optimized for chat, reliable
gpt-4o-mini
OpenAI
Good
Very Fast
$
High volume, cost-effective
gpt-5-nano
OpenAI
Good
Very Fast
$
Budget-conscious, simple queries
gpt-4.1-nano
OpenAI
Good
Very Fast
$
Economical, adequate quality
gemini-2.5-flash-lite
Gemini
Good
Very Fast
$
Most economical Gemini option
gpt-4-turbo
OpenAI
High
Fast
$$
Fast performance, quality responses
gpt-4
OpenAI
High
Medium
$$
Original, reliable, established
claude-3.5-sonnet
Claude (Anthropic)
Highest
Medium
$$$
Complex reasoning, premium support experiences
claude-3.5-haiku
Claude (Anthropic)
High
Fast
$$
Balanced Claude option for most chatbots
claude-3.5-opus
Claude (Anthropic)
Highest
Medium
$$$
Most advanced Claude for critical use cases
claude-3-opus
Claude (Anthropic)
High
Medium
$$$
Earlier flagship Claude, still very capable
claude-3-sonnet
Claude (Anthropic)
High
Fast
$$
General-purpose Claude model
claude-3-haiku
Claude (Anthropic)
Good
Very Fast
$
High-volume, budget-friendly Claude
deepseek-chat
DeepSeek
High
Fast
$
Cost-effective alternative for general chat
For a quick vendor-level comparison between two providers already listed above:
Very strong reasoning, safe and controlled responses
Good quality with very competitive pricing
When to choose
When you want the highest-quality assistant experience and can pay a bit more
When budget is important but you still want strong AI replies
Embedding Models Comparison
Embedding models turn text into searchable vectors. Source material (for example WordPress content, uploads, or pasted text) is split into segments; each segment is embedded. When someone asks a question, the chatbot retrieves the most relevant segments and passes them to the chat model as context.
Model
Provider
Quality
Cost
Best For
text-embedding-3-large
OpenAI
Highest
$$
Maximum search accuracy, large knowledge bases
text-embedding-3-small
OpenAI
High
$
Balanced – recommended for most users
text-embedding-ada-002
OpenAI
Good
$
Legacy option, reliable
gemini-embedding-001
Gemini
High
$
Gemini users, Google ecosystem
Important: Use one embedding model consistently for everything you index in the knowledge base so search stays coherent. Mixing embedding models can reduce accuracy.
AI Provider Comparison
AI Settings also supports Grok (xAI) and OpenRouter. The matrix below focuses on four widely used providers; for Grok or OpenRouter, use the model lists inside Chatbot → AI Settings.
Cost: Stronger chat models usually cost more per message. Embedding cost spikes mainly when you (re)build the knowledge index. Monitor usage and invoices from your AI provider, then adjust models to match quality and budget.
