Australian-made LLM beats OpenAI and Google at legal retrieval

Isaacus, an Australian foundational legal AI startup, has made waves in the AI landscape with the launch of Kanon 2 Embedder, a cutting-edge legal embedding LLM that has outperformed giants like OpenAI and Google in legal information retrieval. Let’s delve into what sets Kanon 2 Embedder apart and why it’s a game-changer in the legal tech domain.

Key Points and Insights:

1. Superior Performance on Massive Legal Embedding Benchmark (MLEB)

Kanon 2 Embedder has achieved remarkable results on the Massive Legal Embedding Benchmark (MLEB), showcasing a 9% higher accuracy than OpenAI Text Embedding 3 Large and a 6% higher accuracy than Google Gemini Embedding. This significant lead in accuracy is coupled with a remarkable speed advantage, running over 30% faster than both competing LLMs.

2. Comprehensive Evaluation Across Multiple Jurisdictions and Domains

The Massive Legal Embedding Benchmark (MLEB) developed by Isaacus evaluates legal information retrieval performance across six jurisdictions, including the US, UK, EU, Australia, Singapore, and Ireland, spanning five domains: cases, statutes, regulations, contracts, and academia. This comprehensive evaluation ensures that Kanon 2 Embedder is versatile and robust in handling diverse legal contexts.

3. Leading the Pack in a Competitive Field

With stiff competition from 20 LLMs, including industry heavyweights like Qwen3 Embedding 8B, IBM Granite Embedding R2, and Microsoft E5 Large Instruct, Kanon 2 Embedder has emerged as the frontrunner on the MLEB, solidifying its position as a trailblazer in legal AI technology. Its innovative approach and superior performance have set a new standard in the field.

Context and Examples:

Imagine a legal research tool that not only provides accurate results but does so swiftly and efficiently. Kanon 2 Embedder’s success on the MLEB underscores its ability to revolutionize legal information retrieval, offering legal professionals a powerful tool to enhance their research and decision-making processes. By surpassing industry giants like OpenAI and Google, Isaacus has demonstrated the strength of Australian innovation in the global AI landscape.

Conclusion and Call-to-Action:

As Isaacus continues to push the boundaries of legal AI with Kanon 2 Embedder, the future of legal information retrieval looks promising. Legal professionals, researchers, and technology enthusiasts can look forward to leveraging this groundbreaking technology to streamline their workflows and gain deeper insights into legal matters. Stay tuned for more updates on the evolving landscape of legal AI and explore how Kanon 2 Embedder can transform the way we approach legal research and analysis.