> ## Documentation Index
> Fetch the complete documentation index at: https://docs.qwairy.co/llms.txt
> Use this file to discover all available pages before exploring further.
# Providers & Credits
> Every AI provider and model Qwairy supports, how each is queried (web scraping vs. official API), and how many credits each response costs.
Qwairy sends your prompts to a wide set of AI providers and models. This page lists every provider and model we support, explains the two ways Qwairy collects their answers, and details how many credits each response costs.
## How Qwairy queries AI models
Qwairy collects answers in **two fundamentally different ways**, and every model falls into one of them. The method determines what you're measuring, where it can be run, and how much it costs.
Qwairy simulates a real user sending your prompt through the model's **public-facing website**, then retrieves the answer straight from that interface.
You see **exactly what a real user sees** in ChatGPT, Perplexity, Gemini, Copilot, Grok, or Google, including AI Overviews and AI Mode. Results can be **geo-localized by country**.
Flat cost: **1 credit** per response.
Qwairy calls the provider's **official API** for a specific, versioned model (e.g. `gpt-5`, `claude-opus-4-8`), with **web search grounding** enabled so answers stay current.
You get a **deterministic, pinned model version**, ideal for comparing a precise model across time or against competitors.
Cost varies by model: **2 to 20 credits** per response.
| | Core models | Premium models |
| -------------------- | ------------------------------------ | ----------------------------------- |
| **Collection** | Scraping of the public web interface | Official provider API |
| **Reflects** | What a real user sees in the app | A specific, pinned model version |
| **Geo-localization** | Yes, by country | Location passed as a hint to search |
| **Web search** | Native to the interface | Enabled via search grounding |
| **Cost** | 1 credit | 2 to 20 credits |
Both methods answer different questions. **Core models** tell you how your brand shows up for a real user right now. **Premium models** let you benchmark a precise model version consistently. Most brands monitor a few Core models first, then add Premium models for deeper competitive analysis.
## Core models: web scraping
All Core models cost **1 credit** per response and reflect the model's live, public web experience.
| Model | Provider | What it is |
| ---------------------- | ------------- | --------------------------------------------------------- |
| **ChatGPT** | OpenAI | The standard ChatGPT web interface |
| **Perplexity** | Perplexity AI | The conversational search-and-answer engine |
| **Copilot** | Microsoft | Microsoft Copilot's web interface |
| **Grok** | xAI | xAI's Grok, via its web interface |
| **Gemini** | Google | Google's Gemini web chat experience |
| **Google AI Overview** | Google | The AI-generated summaries shown in Google Search results |
| **Google AI Mode** | Google | Google Search's conversational AI Mode |
## Premium models: official API
Premium models are queried through each provider's official API. Cost depends on the model.
### OpenAI
| Model | Credits |
| ------------ | ------- |
| GPT-5 Mini | 2 |
| GPT-5 | 4 |
| GPT-5.1 | 4 |
| GPT-5.2 | 5 |
| GPT-5.4 Mini | 3 |
| GPT-5.4 | 6 |
| GPT-5.5 | 11 |
### Anthropic (Claude)
| Model | Credits |
| ----------------- | ------- |
| Claude Haiku 4.5 | 4 |
| Claude Sonnet 4.5 | 12 |
| Claude Sonnet 4.6 | 14 |
| Claude Sonnet 5 | 14 |
| Claude Opus 4.6 | 20 |
| Claude Opus 4.7 | 20 |
| Claude Opus 4.8 | 20 |
### Google Gemini
| Model | Credits |
| ---------------------- | ------- |
| Gemini 2.5 Flash Lite | 3 |
| Gemini 2.5 Flash | 4 |
| Gemini 2.5 Pro | 5 |
| Gemini 3 Flash Preview | 4 |
| Gemini 3.1 Flash Lite | 2 |
| Gemini 3.5 Flash | 3 |
| Gemini 3.1 Pro Preview | 5 |
### Perplexity
| Model | Credits |
| ------------------- | ------- |
| Sonar | 2 |
| Sonar Pro | 3 |
| Sonar Reasoning Pro | 4 |
| Sonar Deep Research | 6 |
### xAI (Grok)
| Model | Credits |
| -------- | ------- |
| Grok 4.3 | 5 |
| Grok 4.5 | 3 |
### Mistral
| Model | Credits |
| -------------- | ------- |
| Mistral Small | 3 |
| Mistral Medium | 4 |
| Mistral Large | 4 |
### DeepSeek
| Model | Credits |
| ----------------- | ------- |
| DeepSeek Chat | 2 |
| DeepSeek Reasoner | 2 |
## How credits work
Qwairy uses a credit-based system. Credits are consumed when generating AI responses, creating content, and importing prompts.
| Operation | Cost | Description |
| --------------------------------- | ----------------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| **Answer generation** | 1 to 20 credits | The base unit: one prompt sent to one model (see tables above). Every total below is a multiple of this. |
| **Brand Perception analysis** | 10 credits per provider | A perception snapshot runs across the available Core providers (up to 6), so up to 60 credits per run. |
| **Content Studio brief** | 50 credits | One article brief. |
| **Prompt import / auto-generate** | \~7 credits per prompt | When you add or import prompts with **auto-generate** enabled, each new prompt is immediately run against your Core models (7 by default). Turn auto-generate off to add prompts without spending credits until you monitor. |
### Calculation examples
**Monitoring run**, 10 prompts x 3 models (ChatGPT Core, GPT-5 Mini, Perplexity Core):
```
10 x (1 + 2 + 1) = 40 credits per run
```
**Prompt import with auto-generate**, 100 prompts run against all 7 Core models:
```
100 x 7 = 700 credits
```
**Brand Perception**, one snapshot across 6 Core providers:
```
6 x 10 = 60 credits
```
**Content Studio**, 1 article brief = **50 credits**.
### Credit balance
View your remaining credits at the top of the dashboard header, or under **Team Management > Usage**.
### Best practices
| Tip | Benefit |
| -------------------------------------- | ---------------------------------------------- |
| Start with Core models | Lower cost, validate your prompts first |
| Use weekly monitoring | Balance data freshness vs. cost |
| Disable auto-generate for bulk imports | Control when credits are spent |
| Add Premium models selectively | Reserve higher-cost models for deep benchmarks |
### Running out of credits
When credits are exhausted, scheduled monitoring pauses and manual generation is blocked. Purchase additional credits to resume.
## Related
* [Monitoring](/documentation/workspace/monitoring): select which models run and set your refresh frequency
* [Core Concepts](/documentation/get-started/core-concepts): the Provider / AI Model vocabulary
* [Billing](/documentation/team/billing): manage your subscription and credits
* [Which AI providers should I prioritize?](/use-cases/setup/which-providers-to-prioritize)