Vane, an open-source AI search tool, lets people choose their model or run one locally instead of relying on a provider’s hosted search service. A MakeUseOf reviewer found that flexibility appealing, but also spent hours wrestling with a self-hosted setup that initially returned no useful search results.
Watch Desk analysis
What happened
Vane, formerly known as Perplexica, uses SearXNG to retrieve web results, then has an AI model produce answers with citations. It offers Speed, Balanced and Quality search modes, and can search the web, discussions or academic papers. The reviewer says it also supports image and video searches and a multi-step research mode.
Users can connect models from OpenAI, Anthropic, Google and Groq with their own API keys, or run models locally through Ollama. The MakeUseOf review describes testing it on a MacBook Air, including a research query that used 11 steps and 45 sources. Those are the reviewer’s results, not a controlled comparison.
Why it matters
The model choice is more than a settings-menu nicety. Local inference can avoid sending queries to a cloud model, while API access lets occasional users pay for usage rather than another monthly subscription. Neither route is automatically free: API costs can accumulate, and local models need suitable hardware.
The reviewer also found the trade-off in day-to-day use. Vane’s search providers initially returned nothing; troubleshooting SearXNG and networking took hours. In one technical test, the reviewer found Perplexity’s answer more useful, despite Vane gathering more sources.
Our read
Vane makes the controls that hosted search often hides available to users willing to manage the plumbing. That could suit people who value model choice, local processing or tinkering. If you want to type a question and get on with your afternoon, the reviewer’s experience suggests the hosted option may still win on convenience. Open source is a meaningful choice, not a magic wand for setup problems.
What to watch
- Whether Vane’s setup and search-provider reliability improve.
- How well its cited answers hold up across more independent testing.
- Whether local models and API usage make the control worthwhile for different users.
Discussion spark: Would you trade the convenience of a hosted AI search service for the choice to run a model locally, or is that control not worth the setup and upkeep?
Sources and evidence
- Perplexity has an open-source alternative now, and I'm hooked – MakeUseOf (19 October 2026, 19:00 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.