Getting Started
Guides
- Prompting & Output Control
- Search & Filtering
- Working with Images
- Integration & Extensibility
Help & Resources
Getting Started
Quickstart
Generate an API key and make your first call in < 3 minutes.
Generating an API Key
Get your Sonar API Key
Navigate to the API Keys tab in the API Portal and generate a new key.
See the API Groups page to set up an API group.
OpenAI SDK Compatible: Perplexity’s API supports the OpenAI Chat Completions format. You can use OpenAI client libraries by pointing to our endpoint. See our OpenAI SDK Guide for examples.
Making Your First API Call
cURL is a command-line tool for making HTTP requests. Set your API key and run the command:
Copy
Ask AI
'export SONAR_API_KEY="your_api_key"
curl --location 'https://api.perplexity.ai/chat/completions' \
--header 'accept: application/json' \
--header 'content-type: application/json' \
--header "Authorization: Bearer $SONAR_API_KEY" \
--data '{
"model": "sonar-pro",
"messages": [
{
"role": "user",
"content": "What are the most popular open-source alternatives to OpenAIs GPT models?"
}
]
}
Here's an example response (raw response at the end)
Here's an example response (raw response at the end)
Response Content
Copy
Ask AI
## 2025 French Open Finals Results
**Men's Singles Final**
- **Champion:** Carlos Alcaraz
- **Runner-up:** Jannik Sinner
- **Score:** 4–6, 6–7^(4–7), 6–4, 7–6^(7–3), 7–6^(10–2)
- **Details:** Carlos Alcaraz successfully defended his title by defeating Jannik Sinner in a dramatic five-set final. The match lasted 5 hours and 29 minutes, making it the longest French Open final in history and the second-longest major final ever. Alcaraz came back from two sets down and saved three championship points, marking a historic comeback. This victory gave Alcaraz his second French Open and fifth Grand Slam title, making him the second-youngest man to win five majors in the Open Era. It was also the first time a French Open singles final was decided by a final-set tiebreak.[1][2]
**Women's Singles Final**
- **Champion:** Coco Gauff
- **Runner-up:** Aryna Sabalenka
- **Score:** (Set scores not fully provided, but Gauff won in three sets)
- **Details:** Coco Gauff rallied from an early deficit to defeat Aryna Sabalenka in a three-set battle. After losing the first set, Gauff fought back to claim her second Grand Slam title, with both coming against Sabalenka (the first being the 2023 US Open). Gauff’s victory marks her first French Open crown at age 21.[5]
**Women's Doubles Final**
- **Champions:** Anna Danilina / Aleksandra Krunić
- **Runners-up:** Sara Errani / (partner not fully specified)
- **Score:** 4–6, 6–1, 6–1
- **Details:** Danilina and Krunić won the women's doubles final in three sets, taking control after dropping the first set.[3]
**Summary Table**
| Event | Winner | Runner-up | Score |
|------------------------|---------------------------------------|----------------------------------|-----------------------------------------------|
| Men's Singles | Carlos Alcaraz | Jannik Sinner | 4–6, 6–7(4–7), 6–4, 7–6(7–3), 7–6(10–2) |
| Women's Singles | Coco Gauff | Aryna Sabalenka | Gauff won in 3 sets (full scores not given) |
| Women's Doubles | Anna Danilina / Aleksandra Krunić | Sara Errani / (unspecified) | 4–6, 6–1, 6–1 |
Alcaraz's men's final win was both historic for its comeback and duration, and Gauff’s victory marked a defining moment in her early career.[1][2][5]
Citations
Copy
Ask AI
[
"https://en.wikipedia.org/wiki/2025_French_Open_%E2%80%93_Men's_singles",
"https://en.wikipedia.org/wiki/2025_French_Open_%E2%80%93_Men's_singles_final",
"https://www.rolandgarros.com/en-us/matches?status=finished",
"https://www.tennis.com/news/articles/who-were-the-winners-and-losers-at-2025-roland-garros",
"https://www.cbssports.com/tennis/news/2025-french-open-results-schedule-as-jannik-sinner-faces-carlos-alcaraz-coco-gauff-earns-first-title/"
]
Sample Search Results (first 2 sources)
Copy
Ask AI
[
{
"title": "2025 French Open \u2013 Men's singles",
"url": "https://en.wikipedia.org/wiki/2025_French_Open_%E2%80%93_Men's_singles",
"date": "2025-04-27",
"last_updated": "2025-06-24"
},
{
"title": "2025 French Open \u2013 Men's singles final",
"url": "https://en.wikipedia.org/wiki/2025_French_Open_%E2%80%93_Men's_singles_final",
"date": "2025-06-08",
"last_updated": "2025-06-24"
}
]
Usage Information
Copy
Ask AI
{
"completion_tokens": 625,
"prompt_tokens": 13,
"total_tokens": 638
}
Raw Response
Copy
Ask AI
{
"id": "d06009f7-06e3-481b-87b9-37878abab471",
"model": "sonar-pro",
"created": 1752790019,
"usage": {
"prompt_tokens": 16,
"completion_tokens": 517,
"total_tokens": 533,
"search_context_size": "low"
},
"citations": [
"https://www.syncfusion.com/blogs/post/best-5-open-source-llms",
"https://www.softkraft.co/gpt3-open-source-alternatives/",
"https://www.semrush.com/blog/chatgpt-alternatives/",
"https://www.datacamp.com/blog/12-gpt4-open-source-alternatives",
"https://zapier.com/blog/chatgpt-alternatives/"
],
"search_results": [
{
"title": "Best 5 Open-Source LLMs for Developers: ChatGPT Alternatives in ...",
"url": "https://www.syncfusion.com/blogs/post/best-5-open-source-llms",
"date": "2025-06-17",
"last_updated": "2025-06-21"
},
{
"title": "6 Proven GPT-3 Open-Source Alternatives [2023 Comparison]",
"url": "https://www.softkraft.co/gpt3-open-source-alternatives/",
"date": "2024-10-11",
"last_updated": "2025-05-16"
},
{
"title": "My 11 Favorite ChatGPT Alternatives for 2025 (Tested Manually)",
"url": "https://www.semrush.com/blog/chatgpt-alternatives/",
"date": "2025-05-22",
"last_updated": "2025-07-13"
},
{
"title": "12 GPT-4 Open-Source Alternatives - DataCamp",
"url": "https://www.datacamp.com/blog/12-gpt4-open-source-alternatives",
"date": "2024-08-01",
"last_updated": "2025-07-04"
},
{
"title": "The 9 best ChatGPT alternatives in 2025 - Zapier",
"url": "https://zapier.com/blog/chatgpt-alternatives/",
"date": "2023-10-18",
"last_updated": "2025-07-16"
}
],
"object": "chat.completion",
"choices": [
{
"index": 0,
"finish_reason": "stop",
"message": {
"role": "assistant",
"content": "The most popular open-source alternatives to OpenAI's GPT models as of 2025 include several robust Large Language Models (LLMs) that have been widely adopted by developers and organizations seeking more control, lower costs, and customization compared to proprietary offerings. The leading models are:\n\n- **Llama 3 & Llama 4 (Meta):** These models from Meta are currently among the most advanced and popular open-source GPT alternatives. Llama 4, released in April 2025, offers significant improvements in speed, multilingual capabilities, and the ability to handle more complex and longer interactions. It is freely available for both research and commercial use and can be integrated into a variety of platforms, including Meta's social applications and custom software[1][3].\n\n- **Mistral:** A fast-growing open-source LLM developed by a startup focused on efficiency and performance, Mistral provides strong performance, especially for deployments requiring smaller model sizes or faster inference[1].\n\n- **Falcon:** Produced by the Technology Innovation Institute, Falcon is recognized for its excellent performance, especially in large-scale languages modeling, and is available under a permissive license for both research and commercial uses[1].\n\n- **BLOOM:** Developed by the BigScience project and coordinated by Hugging Face, BLOOM is a multilingual LLM supporting 46 human languages and 13 programming languages. It is fully open source, intended to democratize access to large language models, and can be freely downloaded for research and commercial purposes[1][2].\n\n- **Pythia:** This suite of models focuses on transparency and reproducibility in LLM research, allowing detailed exploration and fine-tuning of the models[1].\n\nOther noteworthy open-source projects and toolkits include:\n\n- **GPT-NeoX** and **GPT-JT:** Developed by EleutherAI, these models offer architectures and capabilities similar to GPT-3 and are widely used in the open-source AI community[2].\n- **OpenChatKit:** An open-source toolkit for developing ChatGPT-like chatbots, including instructions for training, fine-tuning, and moderating large language models[4].\n- **ColossalChat** and **OPT:** Additional major open-source alternatives that aim to replicate GPT capabilities via community-driven development and transparent licensing[2][4].\n\nThese models empower developers with control over deployment, customization, and data privacy while offering performance competitive with proprietary GPT models, making them top choices for those seeking alternatives to OpenAI's offerings[1][3][2]."
},
"delta": {
"role": "assistant",
"content": ""
}
}
]
}
For a full guide on streaming, including parsing, error handling, citation management, and best practices, see our streaming guide.
Next Steps
Now that you’ve made your first API call, here are some recommended next steps:Models
Explore the different models available and their capabilities.
API Reference
View the complete API documentation with detailed endpoint specifications.
Guides
Read our guides to learn how to get the most out of the Sonar APIs.
Examples
Explore code examples, tutorials, and integration patterns.
Need help? Check out our community for support and discussions with other developers.
Assistant
Responses are generated using AI and may contain mistakes.