14 models on CoreAI can search the live web before answering, so their responses include current events, fresh prices, and today's facts instead of stopping at a training cutoff. Search-capable models come from Sakana, Perplexity, OpenAI, listed newest-first below.
Web-search models cite their sources, which makes them the right choice for research, fact-checking, news summaries, and anything where "as of today" matters. In the CoreAI app these models show a "Search" chip in chat — CoreAI's AI Search tool is built on the same capability.
| Model | Provider | Context | Input $/1M | Output $/1M | Capabilities |
|---|---|---|---|---|---|
| Sakana: Fugu Ultra | Sakana | 1000K | $5 | $30 | Reasoning, Web search, Vision |
| Perplexity: Sonar Pro Search | Perplexity | 200K | $3 | $15 | Reasoning, Web search, Vision |
| OpenAI: GPT-4o-mini Search Preview | OpenAI | 128K | $0.15 | $0.60 | Web search |
| OpenAI: GPT-4o Search Preview | OpenAI | 128K | $2.50 | $10 | Web search |
| Perplexity: Sonar Reasoning Pro | Perplexity | 128K | $2 | $8 | Reasoning, Web search, Vision |
| Perplexity: Sonar Pro | Perplexity | 200K | $3 | $15 | Web search, Vision |
| Perplexity: Sonar Deep Research | Perplexity | 128K | $2 | $8 | Reasoning, Web search |
| Perplexity: Sonar | Perplexity | 127K | $1 | $1 | Web search, Vision |
| OpenAI: GPT-4o (2024-11-20) | OpenAI | 128K | $2.50 | $10 | Web search, Vision |
| OpenAI: GPT-4o (2024-08-06) | OpenAI | 128K | $2.50 | $10 | Web search, Vision |
| OpenAI: GPT-4o-mini | OpenAI | 128K | $0.15 | $0.60 | Web search, Vision |
| OpenAI: GPT-4o-mini (2024-07-18) | OpenAI | 128K | $0.15 | $0.60 | Web search, Vision |
| OpenAI: GPT-4o | OpenAI | 128K | $2.50 | $10 | Web search, Vision |
| OpenAI: GPT-4o (2024-05-13) | OpenAI | 128K | $5 | $15 | Web search, Vision |
Most can't — a standard LLM only knows its training data. The 14 models on this page are the exception: they run a live web search as part of answering, then ground the response in what they found, usually with citations.
Anything time-sensitive: news, market prices, sports results, product research, travel planning, and fact-checking claims. For timeless tasks like writing or coding, a standard model is faster.
Yes — responses typically include links to the pages the model consulted, so you can verify every claim. CoreAI displays these citations directly in the chat.
It's a layer on top of one. The model runs searches, reads the top results, and synthesizes a direct answer — you get the conclusion with sources, instead of a page of links to read yourself.
Chat with GPT-5, Claude, Gemini and hundreds more — all in one app.