Announcing Tabnine Protected 2: A license-safe LLM that performs as strong as the best

Generative AI tools face adoption challenges in large organizations due to concerns over copyright and license compliance, as evidenced by the hesitance of many CIOs and ongoing lawsuits about the use of content in training models without explicit permission. Tabnine addresses these concerns with its proprietary models, particularly the Tabnine Protected, which is trained on permissively licensed code to avoid legal liabilities. The latest iteration, Tabnine Protected 2, offers enhanced performance, surpassing models like GPT-3.5 Turbo, while maintaining its focus on privacy and compliance. It supports over 600 programming languages and employs retrieval-augmented generation (RAG) to improve context understanding and personalization. Evaluations using benchmarks such as HumanEval and MultiPL-E demonstrate its superior code generation capabilities, and real-world customer feedback shows higher acceptance rates compared to other models. Tabnine ensures no data retention, offering peace of mind with IP indemnification and deployment flexibility, allowing users to switch between models based on project requirements.