#USA #San Francisco #Aurora #LILT #multilingual AI

LILT Launches AURORA: Revolutionizing Multilingual AI Performance Measurement

On September 30, 2026, LILT took a significant step in the field of multilingual AI with the launch of AURORA, the first-ever Multilingual AI Leaderboard designed to evaluate cutting-edge models on non-English tasks. As global enterprises expand their operations, it is essential to address language diversity, especially considering that many existing benchmarks have historically focused on English-centric evaluations.

Meeting the Multilingual Challenge

In today’s interconnected world, businesses are increasingly using AI agents across various languages and cultures. The traditional approach of relying on English-based metrics often led to gaps in representation and performance assessment. Spence Green, CEO and co-founder of LILT, emphasizes the importance of this need: “The industry is choosing models on a scoreboard that stops at English. That gap used to cost you an awkward translation. Now that agents are taking real actions in the real world, it costs you a wrong decision, and you find out from your users.”

AURORA seeks to bridge this gap by providing insights into multilingual AI performance through scientifically validated benchmarks designed by native-language domain experts. By moving beyond English-centric models, AURORA allows businesses to make informed decisions based on accurate assessments tailored to local users.

The AURORA Benchmark Suite

The leaderboard features a suite of benchmarks that cover various enterprise applications, ensuring a comprehensive evaluation of performance in multilingual environments. Some key components of the AURORA benchmark suite include:

• - Multilingual Terminal-bench: Challenging coding tasks specifically created for software developed for native-language users.
• - Multilingual τ³-bench: Simulating multi-turn customer support scenarios in industries such as airlines, telecommunications, retail, and banking.
• - Multilingual MultiChallenge: Evaluating long-context instruction-following, memory, and self-coherence in conversations.
• - Multilingual GAIA-v2-LILT: Focusing on agentic reasoning and the effective use of tools in practical scenarios.

Each of these benchmarks is designed to authentically represent real-world tasks faced by businesses in multilingual contexts, making AURORA an indispensable tool for enterprises seeking to optimize their AI strategies.

The Importance of Accurate Measurement

Recent analyses conducted by LILT's PhD-led Applied AI practice have revealed the considerable variability in model performance across different languages. For instance, in coding tasks, the performance of GPT 5.5 is notably highest in Spanish, while Claude Opus 5.5 outperforms in Japanese. Muse Spark 1.3 leads in Serbian. This variation underscores the necessity for businesses to utilize specialized tools like AURORA that take into account the distinct nuances and challenges tied to each language.

Accessing AURORA

For those interested in exploring the AURORA leaderboard, it is now live and accessible at https://aurora.lilt.com/. Users can compare AI models based on various languages and tasks to find the most suitable solutions for their specific needs. Additionally, AI teams looking for customized benchmarks can reach out to LILT for tailored evaluations.

About LILT

LILT is renowned for its commitment to advancing agentic AI, allowing enterprises, public sector organizations, and frontier labs to streamline multilingual capabilities at scale. With a robust portfolio featuring applied AI services, LILT provides private multilingual benchmarks and evaluation resources essential for developing superior models. The company partners with prominent organizations such as NVIDIA, Intel, and L'Oréal to expand global outreach.

In conclusion, the launch of AURORA represents not just an evolution in AI measurement but a necessity for genuinely international businesses. By prioritizing cultural and linguistic diversity in AI assessments, LILT is paving the way for a more equitable technology landscape that acknowledges and harnesses the power of multilingualism.