Keenable Team
Keenable ranks #1 on the Artificial Analysis Search Index
First and second of 26 configurations: both modes of the Keenable Web Search API top the Artificial Analysis Search Index, ahead of Exa and Perplexity.
Artificial Analysis has benchmarked both modes of the Keenable Web Search API. With GPT-5.6 Luna as the answer model, Keenable Pro scores 83 and Keenable Realtime 82, first and second of 26 configurations. Exa (auto, dynamic highlights) and Perplexity (medium) follow at 80.

On the Search Index vs. cost per task chart, Keenable Pro is the top point of the Pareto frontier: no configuration scores higher at any price, and the closest competitors, Exa (auto, dynamic highlights) and Perplexity (medium) at 80, cost $0.064 and $0.070 per task.

For third-party search APIs, the benchmark holds the answer model and the agent harness constant and changes only the search API, so the gap between scores is the gap between search tools.
GPT-5.6 Luna scores 33 alone and 83 with Keenable
Without search, GPT-5.6 Luna scores 33 on the index. Keenable Pro takes it to 83, a 50-point lift. Exa (auto, dynamic highlights) and Perplexity (medium) add 47 points each.

Cost per task: $0.040 per task against a $0.063 median
A full task, model inference included, costs about $0.040 with Keenable, against the board median of $0.063 that Artificial Analysis reported on November 10.

Time per task: about 15 seconds
Tasks also finish faster: 15.4 seconds in Realtime mode and 16.0 seconds in Pro, against 36.0 seconds for Exa (auto, dynamic highlights) and 30.2 seconds for Perplexity (medium).

How the Search Index is scored: 1,700 tasks, three benchmarks
The Search Index is the average of three benchmarks, weighted equally.
DeepSearchQA
900 research questions where the answer is a list, scored on how complete and accurate that list is. Both Keenable modes score 88, the top two results on the board, ahead of Exa (auto, dynamic highlights) at 86 and Perplexity (medium) at 81.

BrowseComp
200 hard questions that take several searches to answer, where only the exact answer counts. The model alone scores 17. With Keenable Pro it scores 87, level with Perplexity (medium) for first place, ahead of Exa at 83.

AA-Omniscience
600 factual questions across six domains, kept private by Artificial Analysis. The model alone scores 38. With Keenable it scores 73, level with Exa.

Read the Artificial Analysis methodology for the full evaluation setup and scoring rules.
Test the Keenable Web Search API on your own queries
Keenable is available through our Search API and MCP. Run it in your own agent and compare answer quality, total task cost, and completion time on the tasks your users care about.
Documentation | Create an account
Results from the Artificial Analysis leaderboard as of November 10, 2026. Keenable has been evaluated with GPT-5.6 Luna (medium).



