Leni Achieves Outstanding Results on Crucial AI Performance Benchmarks

Leni Surpasses Major AI Benchmarks

Leni, an innovative AI-powered analytics platform catering to the commercial real estate sector, has recently made headlines by excelling in four significant independent AI benchmarks. This performance places Leni ahead of established AI platforms from industry giants like OpenAI, Anthropic, Google, and Perplexity, demonstrating its prowess and reliability in delivering accurate analytics.

In the DRACO Benchmark, a test for deep research capabilities designed by Perplexity AI in collaboration with Harvard, Leni achieved an impressive score of 71.6%. This bested the equivalent products from its notable competitors, underscoring Leni's ability to produce research that senior analysts in the field can trust and endorse. Furthermore, the platform ranked among the top two globally in the SpreadsheetBench Verified assessments, successfully completing 365 out of 400 real spreadsheet tasks, again indicating its effectiveness in practical applications.

On the BullshitBench, a benchmark that measures how well AI can resist nonsensical prompts, Leni demonstrated exceptional accuracy by identifying fabricated premises 98% of the time—performing better than all 142 other public AI models. This performance not only illustrates Leni's advanced understanding and filtering of complex queries but also its commitment to providing reliable outputs in scenarios where accuracy is essential, such as commercial real estate.

Leni also made strides on the GAIA benchmark, jointly developed by Meta and HuggingFace, which assesses the system’s capability to execute multi-step tasks without deviating off-track. Leni scored 77.0% on their validation set, significantly surpassing competitors like Genspark, Manus, and OpenAI Deep Research. This capacity to handle intricate, real-world tasks echoes the very needs of commercial real estate, where precision can significantly impact financial dealings.

According to Arunabh Dastidar, the CEO and Co-Founder of Leni, the current challenge within AI adoption lies not with the models themselves but with the architecture or