XNews

Keenable Search 1.0 Tops Artificial Analysis Search Index with Pro and Realtime Modes

Keenable Search 1.0 Tops Artificial Analysis Search Index with Pro and Realtime Modes

Keenable Search 1.0 ranked first with its pro mode scoring 83 and second with its realtime mode scoring 82 on the Artificial Analysis Search Index, ahead of Exa and Perplexity at 80. Benchmarked across 1,700 search-focused tasks in the Stirrup agent harness, both modes scored 88 F1 on DeepSearchQA and cost about $0.040 per task at $4 per 1,000 searches, over a third below the board median.

Event version 1.

Key insights

Top index scores

Keenable Search 1.0 pro scored 83 and realtime scored 82 on the Artificial Analysis Search Index, up from 67 for both modes in August. Both modes lead Exa (auto, dynamic highlights) and Perplexity (medium) at 80.

Strong DeepSearchQA results

Both Keenable modes scored 88 F1 on DeepSearchQA, ranking first and second of 26 ahead of Exa at 86. Pro also ranked first on BrowseComp at 87% and on AA-Omniscience at 73%.

Low cost per task

At $4 per 1,000 searches, Keenable costs about $0.040 per task in both modes, over a third below the board median of $0.063.

Fast search times

Realtime mode took 2.3 seconds of search time per task and pro took 3.0 seconds, against a board median of 12.5 seconds. Each search returned in about 330ms (realtime) and 390ms (pro) on average.

Benchmark setup

Testing used GPT-5.6 Luna (medium) in the Stirrup agent harness with the Keenable Search API in realtime and pro modes across 1,700 search-focused tasks. Keenable's search cards were disabled for testing.

Source posts