Overview
WPiko AI Chatbot uses OpenAI’s Responses API. You choose the model and how much it should “think” in WPiko AI Chatbot → AI & Training → AI Settings → Model & Knowledge → AI model settings. The WordPress sidebar’s Settings shortcut opens the same screen. The model you pick decides answer quality, speed and cost.
Available Models
Models are listed from newest to oldest. Prices shown are OpenAI’s standard rates per 1 million tokens (input / output), as shown in the plugin. Some models charge more for long prompts. Check OpenAI’s pricing page for the latest rates.
| Model | Best for | Context | Price (in / out) |
|---|---|---|---|
| GPT-6 Astra | OpenAI’s most capable model for complex reasoning and research | 1.05M | $10.00 / $50.00 |
| GPT-6.1 Sol | Near-Astra quality for professional work at a lower cost | 1.05M | $2.00 / $10.00 |
| GPT-6.1 Terra | Fast visual reasoning and clear, natural writing | 1.05M | $1.00 / $5.00 |
| GPT-6 Sol | Complex and agentic tasks | 1.05M | $2.00 / $10.00 |
| GPT-6 Luna (default) | Efficient, high-volume tasks — a great choice for most chatbots | 1.05M | $0.10 / $0.50 |
| GPT-5.6 Sol | Frontier GPT-5.6 model for complex work | 1.05M | $4.00 / $20.00 |
| GPT-5.6 Terra | Balance of intelligence and cost | 1.05M | $2.00 / $12.00 |
| GPT-5.6 Luna | Cost-sensitive, high-volume workloads | 1.05M | $0.20 / $1.20 |
| GPT-5.5 | Advanced reasoning and professional workflows | 1.05M | $5.00 / $30.00 |
| GPT-5.4 | Agentic and professional workflows | 1.05M | $2.50 / $15.00 |
| GPT-5.4 Mini | Strong, lower-cost GPT-5.4 model | 400K | $0.75 / $4.50 |
| GPT-5.2 | Coding and agentic tasks | 400K | $1.75 / $14.00 |
| GPT-4.1 | Smartest non-reasoning model | 1.05M | $2.00 / $8.00 |
| GPT-4.1 Mini | Smaller, faster GPT-4.1 | 1.05M | $0.40 / $1.60 |
Default model: new installations start with GPT-6 Luna. Updating the plugin keeps the model you already chose.
Removed models: GPT-5, GPT-5 Mini, GPT-5 Nano, GPT-5.1 and GPT-5.4 Nano were removed in version 2.1.0 because OpenAI deprecated them. If one of these was selected, the chatbot automatically uses GPT-6 Luna instead. Open AI Settings → Model & Knowledge and save a supported model to make the choice explicit.
“Your OpenAI account cannot use this model”: some newer models require organization verification in your OpenAI account. Verify your organization in OpenAI, or choose a different model.
Model Settings
Reasoning Effort
Controls how much the model thinks before answering. More effort can give better answers to complex questions but makes replies slower and uses more tokens. The options shown depend on the model:
| Model | Reasoning options | Default |
|---|---|---|
| GPT-6 Astra, GPT-6.1 Sol | Low, Medium, High, X-High, Max | Medium |
| GPT-6.1 Terra, GPT-6 Sol, GPT-6 Luna, GPT-5.6 Sol / Terra / Luna | None, Low, Medium, High, X-High, Max | Medium |
| GPT-5.5, GPT-5.4, GPT-5.4 Mini, GPT-5.2 | None, Low, Medium, High, X-High | None |
| GPT-4.1, GPT-4.1 Mini | Not available (non-reasoning models) | — |
- None: fastest replies, no extra thinking.
- Low / Medium: good for most customer-support chatbots.
- High / X-High / Max: deepest thinking for difficult questions. Slower and more expensive.
The old Minimal option was removed because it conflicts with file search. Saved “Minimal” settings automatically use Low.
Verbosity
Controls how long and detailed answers are (GPT-5 and newer models): Low for short answers, Medium (default) for balanced answers, High for detailed explanations.
Streaming Replies
Replies appear word by word as they are generated, so visitors don’t wait for the full answer. Streaming is on by default. If a browser doesn’t support it, the chatbot falls back to showing the full reply at once. Developers can turn streaming off with the wpiko_chatbot_streaming_enabled filter.
Models Used by Helper Features
Some features use their own model, regardless of the chatbot model you select:
- Conversation translation (Conversations page): GPT-6 Luna
- Contact message enhancement (Pro): GPT-6 Luna
- Scan Website Q&A generation (Pro): GPT-6.1 Terra with Medium reasoning and Medium verbosity
Developers can change these with the wpiko_chatbot_ai_feature_config filter.
Choosing a Model
- Most websites: GPT-6 Luna with Low or Medium reasoning. It’s fast and very affordable.
- Complex products or technical support: GPT-6.1 Sol or GPT-6.1 Terra.
- Highest quality regardless of cost: GPT-6 Astra.
- Fastest replies: GPT-6 Luna with reasoning set to None.
Test a few questions after changing the model, and set User Limits to keep costs predictable.
Troubleshooting
- Slow replies: lower the reasoning effort or choose a faster model.
- Weak answers: try a more capable model, review your System Instructions, and make sure the chatbot has learned your content (Quick learn, files or Scan Website).
- The model isn’t in the list any more: it was deprecated by OpenAI. The chatbot uses GPT-6 Luna until you pick another model.
- “Cannot use the selected model”: verify your OpenAI organization or choose another model.
