LILT Announces V3.5 Translation Models with Improved Quality

Newest Models Outperform Market Peer Models in Inference Time and Quality

• • •

San Francisco, CA - April 4, 2024 - LILT, the leading AI solution for enterprise translation and content creation, today announced the platform’s newest models, security capabilities, and API enhancements. The 90 minute live virtual event focused heavily on LILT-V3.5, LILT’s newest models that leverage reinforcement learning to optimize for content accuracy and fluency.

“With LLMs powering our platform at its core, model performance is critical for our business and our customers,” said LILT CEO Spence Green. “We’re excited to share that we continue to provide industry-leading translation models with best-in-class accuracy and latency.”

LILT-V3.5 Models

LILT’s newest-generation LLMs offer improved speed and accuracy, outperforming popular general purpose and translation models including GPT-4, DeepL, and ModernMT. LILT-V3.5 has 6X more parameters than LILT-V2, with inference improvements that have reduced translation times. LILT-V3.5 is based on the transformer architecture, featuring optimized fused attention operators and direct preference optimization. Purpose-built for translation, LILT-V3.5 translation models are optimized for translation tasks including tag projection, terminology integration, document structure, and other localization use cases.

On-Premise Security Compliance

LILT’s newest integration with Palantir FedStart allows multiple customers to access LILT on premise in an IL4/IL5 environment, enabling the use of LILT for processing of content that is controlled, has higher sensitivity, and is mission critical. With the ability to newly deploy LILT across additional agencies, teams, and environments, government analysts, translators, and project managers working on GovCloud IL4 environments can now process more multilingual content than ever before, utilizing LILT software to efficiently translate material.

API Enhancements

LILT released a modernized, well-documented, and extensively tested public API that caters to both current and future LILT customers. The new API is designed for developers and organizations seeking a seamless, flexible, and secure way to self-configure integration with LILT. This capability ensures continuity for existing users while offering new, customizable options through API gateways. LILT released these API upgrades in tandem with additional improvements to its low-code/no code Connector self-orchestration feature, further enabling businesses to self-serve integration setup and automated content exchange.

These new features are available to enterprise and public sector LILT customers and are in production starting today.

For more information about LILT and its AI-powered solutions, please reach out to contact@lilt.com.

• • •

About LILT

LILT is the leading AI solution for enterprise translations. Our stack, made up of our Contextual AI Engine, Connector APIs, and human feedback, enables global organizations to adopt a true AI translation strategy, focusing on business outcomes instead of outputs. With LILT, innovative, category-defining organizations like Intel, ASICS, WalkMe, and Canva are using AI technology to deliver multilingual, digital customer experiences at scale.

Visit us at lilt.com or contact us at contact@lilt.com.