LeXi AI Achieves Highest Score on India’s Bar Exam Benchmark
On July 14, 2026, LeXi AI announced from New Delhi that its legal‑focused artificial‑intelligence platform attained a 98.95% accuracy rate – 94 correct answers out of 95 questions – on the All India Bar Examination (AIBE) 2020 paper. This performance placed LeXi AI ahead of four leading general‑purpose models: GPT‑5.5 (96.84% accuracy with 92 correct answers and three incorrect), Gemini 3.1 Pro (95.79% accuracy with 91 correct, three incorrect and one unanswered), Claude Opus 4.8 (90.53% accuracy with 86 correct and nine incorrect), and DeepSeek v3.2 (89.47% accuracy with 85 correct and ten incorrect).
The AIBE, administered by the Bar Council of India, is the mandatory qualifying examination for aspiring advocates. The official question paper and answer key are publicly available, enabling any party to replicate the test. In this evaluation, all five models received the identical AIBE 2020 paper under the same conditions, without any specialised prompting, and each response was scored against the Bar Council’s published key. LeXi AI recorded one incorrect answer and no unanswered questions; the other models had varying numbers of incorrect and, in Gemini’s case, one unanswered question.
LeXi AI’s founder and CEO, Onkar Rana, emphasized that the benchmark illustrates the growing importance of vertical AI solutions that are engineered to master a single professional domain rather than to be generalists. He noted that lawyers require technology that comprehends the language, structure, and statutory nuances of Indian law, and that the public, reproducible nature of the AIBE test provides transparent evidence of technical capability.
LeXi AI describes its platform as a tool for legal research, drafting, and statutory interpretation serving lawyers, law firms, institutions, and public bodies within the Indian legal system. The press release was issued under an arrangement with PNN, with PTI stating it takes no editorial responsibility for the content.