- Tabnine launches Tabnine Protected 2, a new large language model focused on performance and compliance.
- Model trained exclusively on permissively licensed code to eliminate legal risks and ensure data privacy.
- Outperforms previous models, including GPT-3.5 Turbo, in terms of performance and accuracy.
- Features include zero data retention, broad language support (over 600 languages), and advanced retrieval-augmented generation (RAG) techniques.
- Available to all current Tabnine users at no extra cost; new users can access a 90-day free trial.
Main AI News:
Tabnine, a pioneer in AI code assistance, has unveiled Tabnine Protected 2, an enhanced large language model (LLM) that ensures superior performance while upholding rigorous compliance with copyright and licensing standards. As generative AI technology continues to evolve, concerns about legal and privacy issues have hindered its widespread adoption. Recent studies indicate that a significant number of CIOs remain cautious due to ongoing legal uncertainties surrounding unlicensed training data.
Tabnine has been a leader in addressing these concerns by offering models trained exclusively on high-quality, permissively licensed code. Tabnine Protected 2 takes this commitment further, providing a model trained solely on permissively licensed data to eliminate legal risks and uphold data privacy. The introduction of IP indemnification further mitigates potential legal challenges.
Historically, models with restricted training data often faced performance trade-offs compared to their unrestricted counterparts. However, Tabnine Protected 2 has effectively bridged this gap, delivering performance that rivals and surpasses models trained on extensive data sets, such as GPT-3.5 Turbo.
Peter Guagenti, President of Tabnine, emphasized the significance of this launch: “With Tabnine Protected 2, we are redefining the benchmarks for AI performance and compliance in software development. The era of hesitation is over. Our clients can now benefit from top-tier AI performance while adhering to strict corporate standards and enjoying full data privacy.“
Developers using Tabnine can now select from various LLMs to power their AI development tools, with the flexibility to switch models as needed. This versatility, coupled with Tabnine’s commitment to zero data retention, positions it as the premier choice for developers seeking comprehensive control and privacy.
Tabnine Protected 2 Highlights:
- Enhanced Performance and Accuracy: Benchmarked against HumanEval and MultiPL-E, the new model achieves superior scores and demonstrates immediate user value.
- Robust Data Privacy: Deployable in any setting, including on-premises, VPC, secure SaaS, or air-gapped environments.
- Advanced Contextual Performance: Utilizes cutting-edge retrieval-augmented generation (RAG) for more relevant and tailored responses.
- Broad Language Support: Expands support to over 600 programming languages and frameworks, surpassing the previous model’s capabilities.
- Commitment to Privacy: Ensures zero data retention, protecting customer code from being stored or used in public model training.
Tabnine Protected 2 is available to all current Tabnine users at no extra charge. Developers can access the new model by updating their IDE plugins. New users can start with a 90-day free trial of Tabnine Pro.
Conclusion:
Tabnine Protected 2 sets a new benchmark in AI-driven software development by addressing critical concerns related to copyright compliance and data privacy. This model’s advanced features and performance enhancements make it a compelling choice for developers who need robust, legally compliant AI tools. As AI adoption continues to grow, Tabnine’s focus on legal and privacy standards could influence the broader market, encouraging other providers to prioritize similar measures. This development underscores the increasing importance of balancing high performance with stringent compliance in the AI industry.
