Open-source agent harness providing uniform tools, sandbox and turn budgets for MKA evaluation.
Google and evaluation firm Artificial Analysis released MKA, a benchmark of 200 multimodal knowledge-work tasks spanning 10 industries that mirror 3-8 hours of expert human effort. Gemini 4 Carbon placed second while costing roughly one-sixth of the top model per task, and the benchmark is now open to developers via Artificial Analysis.
Keenable Search 1.0 ranks first in pro mode with 83 and second in realtime mode with 82 on the Artificial Analysis Search Index, both ahead of Exa and Perplexity at 80. Priced at $4 per 1k searches, it costs about $0.040 per task, over a third below the board median, while completing tasks in roughly 15 seconds for 2nd and 3rd place on speed. Both modes improved about 15 points from the 67 scored in the August benchmark.